The Fork and Pull-Request model is widely used in collaborative development of open source software (OSS), fostering innovation through independent repository copies, but it can also lead to inefficiencies and fragmentation. A key underexplored aspect is the integration effectiveness—the degree to which distributed original commits across forks are effectively integrated back into the main repository. It plays a critical role in OSS project productivity but remains poorly understood. In response, we introduce convergence entropy, a novel metric that quantifies the integration effectiveness by measuring the similarity between distributions of original and merged commits across forks, adjusted for integration ratio. This metric highlights not only the volume of contributions but also their diversity and coordination, offering a unique lens to understand forking practices. Moreover, we explore the relationship between convergence entropy and three dimensions of OSS project productivity, showing significant correlations. We also observe that other factors can alter this dynamic.

Multimodal deepfake technologies have emerged rapidly in recent years, with wide application prospects in various fields. The conventional single-training paradigm with inherent limited generalization illustrates inadequate for addressing the continuous evolution of multimodal deepfakes. However, fine-tuning a model with new deepfake data faces past forgery patterns loss and the significant domain shift in diverse novel multimodal deepfake technologies. To address these issues, we propose a novel Quantum State Analytic Incremental Adaptability Learning method (Qsaint) for multimodal deepfake detection. To stabilize prior deepfake memory, Qsaint recursively learns detection-label mapping relations for the new deepfakes artifact with a closed-form solution, preserving the distribution memory from the historical deepfake domains without accessing previous videos. During incremental learning stages, we propose a deepfake quantum state adaptability module inspired by quantum information science. It adapts to the new forgery states and aligns them with the historical deepfake knowledge through cooling and evolution operations, eliminating deepfake domain shift issues. Comprehensive experiments demonstrate that Qsaint significantly mitigates the memory interference of historical deepfakes, effectively balancing the adaptability for new forgery tasks with the memorization of known deepfake patterns.

Multi-view clustering (MVC), as an important machine learning task, aims to group data into distinct groups by leveraging complementary and consistent information across multiple views. During the last two decades, it has been widely studied, and many methods have been proposed, which has brought incredible development to this field. However, few works comprehensively summarize existing methods and point out the potential challenges in this field for the next decades. To this end, our survey thoroughly reviews existing MVC methods according to three taxonomies, i.e., techniques, fusion strategies, and scenarios. Specifically, seven typical techniques, four fusion strategies, and five typical scenarios are included. Besides, we also collect the commonly used datasets and analyze the performance of typical MVC methods. Moreover, we summarize six application scenarios of existing MVC methods ranging from computer vision, and information retrieval tasks to medical diagnosis and bio-informatics. In particular, we point out seven interesting future directions in this field, which will definitely enlighten the readers.

Exemplar-free Class-Incremental Learning (EFCIL) poses a significant challenge in mitigating catastrophic forgetting, due to the absence of exemplars. Recently, analytic learning-based methods propose a recursive alignment procedure to execute EFCIL in a phase-invariant manner and show state-of-the-art performance. However, they heavily rely on a frozen feature extractor trained with the initial dataset to avoid the misalignment between feature and label spaces, ignoring the importance of acquiring generalizable features across incremental tasks for performance improvement. To tackle this, we rethink the obscured sub-optimality of analytic learning-based methods, particularly through empirical reevaluation, and then introduce the Multi-head analytic learning (Muheal) approach. Muheal forms the multi-head model with a delicate feature extractor, thereby introducing a feature optimization procedure and a forgetting compensation module to balance the learning and forgetting. Specifically, within the feature optimization procedure, the feature extractor seeks to learn more generalizable features in a self-supervised manner using the fully-connected classification head. An analytic learning-based classification head follows to align the feature-label space. Additionally, we employ the compensation module to generate and align pseudo-features with a replicated analytic head, thus preventing overfitting and testing. Comprehensive experiments on several benchmark datasets have demonstrated that Muheal significantly outperforms existing state-of-the-art EFCIL methods and is comparable, if not superior, to methods that use replay techniques.

Active Learning (AL) aims to reduce data annotation costs by selecting the most informative samples from an unlabeled data pool. Traditional AL methods often rely on a single snapshot to identify uncertain or representative samples, often overlooking the poor generalization of a single model. Recent AL studies have attempted to address this issue by tracking a broader range of training dynamics for data selection, typically using averaging or accumulating manner. However, both our theoretical and experimental analyses reveal that these methods obscure the variability inherent in the training process, potentially prioritizing hard-to-learn samples that result in poor generalization. In this paper, we propose a novel AL method termed as Dynamic Confidence Variance (DCoV), that seamlessly integrates variability with the training dynamic to effectively identify a well-generalized Coreset. DCoV leverages the variance of the model’s prediction confidence throughout the training process for active sampling and model training. Our theoretical analysis demonstrates that DCoV provides a lower bound on the population risk of the model learned from selected labeled subset, spanning the entire training process. Extensive experiments demonstrate that our approach significantly outperforms existing state-of-the-art AL methods on various balanced and imbalanced benchmark datasets across various modalities.

In recent years, blockchain has been increasingly applied to authoritative institutions (i.e., authoritative blockchains) to strengthen their authority and reputation by providing reliable and secure data to increase transparency, reducing fraud, and enhancing efficiency for distributed applications (DApps) like electronics certificate, land registration, and e-voting, etc. Blockchain systems in these scenarios often have the features of small node-size, high node-reputation and high node-performance, e.g., government blockchains. However, as one of the core technologies of blockchain, distributed consensus protocols are often designed for large-scale business DApps; their efficiency can be further improved when applied to authoritative blockchains. In this paper, taking into account the essential features of blockchain applications for institutions like government departments, we propose a consensus protocol known as Flying Blocks (FB) to further enhances the efficiency and practicality. FB combines the advantages of the delayed confirmation from Nakamoto consensus with the traditional voting-based BFT consensus to simplify the consensus process and reduce the network resources consumption. To evaluate the protocol's performance, we conduct comparison experiments with Hotstuff and RAFT. The results demonstrate that FB outperforms Hotstuff. Furthermore, to the best of our knowledge, FB is the first BFT protocol that outperforms RAFT, a CFT consensus protocol, deployed in Blockchain systems in terms of transaction processing efficiency.

The advent of advanced technologies, such as large models, has precipitated a surging demand for computational resources, thereby driving the evolution of cloud computing services from single-cloud architectures to multi-cloud paradigms. The accompanying question is how to enable resources from different cloud service providers to collaborate efficiently. Although existing works have explored multi-cloud scheduling methods, most of these works are centralized scheduling, where the decision-maker can schedule computing resources across all clouds. However, this is nearly impossible given the current situation in which computing resources in different clouds come from various cloud service providers. In response to this challenge, JointCloud, a multi-cloud cooperation architecture, has been proposed, which aims at enhancing the cooperation among multiple clouds to provide efficient multi-cluster services. Following the idea of JointCloud, proposes a multi-objective evolutionary algorithm (MOEA) based method for task scheduling in multi-cluster cloud computing environments, without intervening in the intra-cluster scheduling scheme. In the proposed method, we construct a mathematical model with the optimization objectives of minimizing overall waiting time and load imbalance between clusters based on the actual operation data in China Computing NET (C$^{2}$NET). In addition, we also develop an MOEA specifically tailored to address this problem. The performance of the proposed MOEA and existing state-of-the-art MOEAs is examined on the proposed problems. Comparison results highlight the promising performance of the proposed MOEA, the specifically tailored algorithm in effectively addressing the multi-cluster task scheduling problem. In addition, we also compared the results of the MOEAs with the results of three classical scheduling methods, the results proved the effectiveness of the MOEA-based method on this problem.

As a kind of infrastructure-as-code, Dockerfile specifies the structure and functionality of a built Docker image and thus plays an important role in the containerized software development process. Nowadays developers need to spend extra time and effort configuring their Dockerfiles in addition to their regular coding work, which requires knowledge and skills orthogonal to those entailed in other software-related experiences. Poorly written Dockerfile code often introduces errors and maintenance costs. However, little automated support is available for assisting developers in configuring Dockerfiles. In this study, we first conduct an online survey to investigate Docker developers’ perceptions of Dockerfile writing, highlighting the needs and potential benefits of Dockerfile auto-completion techniques. Then, we introduceDOCKERFILL, a pre-trained model based approach that provides completion suggestions for Dockerfile-specific code.DOCKERFILLleverages multi-layer Transformer architecture with syntax-aware multi-task learning, which includes contextual file information and three pre-training tasks, i.e., masked language modeling, syntax type identification, and masked identifier prediction. To evaluateDOCKERFILL’s effectiveness, we collect a dataset of 6,350 high-quality real-world Dockerfiles. Our empirical results show that DOCKERFILL provides up to 52.38% accuracy for token-level completion and 19.69% exact match for line-level completion, outperforming the baselines by 7.32%-37.67% and 1.97%-19.69%, respectively. Also,DOCKERFILLobtains significantly higher human evaluation scores compared to the baselines.

Logit-based knowledge distillation (KD) is commonly used to mitigate catastrophic forgetting in class-incremental learning (CIL) caused by data distribution shifts. However, the strict match of logit values between student and teacher models conflicts with the cross-entropy (CE) loss objective of learning new classes, leading to significant recency bias (i.e. unfairness). To address this issue, we rethink the overlooked limitations of KD-based methods through empirical analysis. Inspired by our findings, we introduce a plug-and-play pre-process method that normalizes the logits of both the student and teacher across all classes, rather than just the old classes, before distillation. This approach allows the student to focus on both old and new classes, capturing intrinsic inter-class relations from the teacher. By doing so, our method avoids the inherent conflict between KD and CE, maintaining fairness between old and new classes. Additionally, recognizing that overconfident teacher predictions can hinder the transfer of inter-class relations (i.e., dark knowledge), we extend our method to capture intra-class relations among different instances, ensuring fairness within old classes. Our method integrates seamlessly with existing logit-based KD approaches, consistently enhancing their performance across multiple CIL benchmarks without incurring additional training costs.

Agents significantly enhance the capabilities of standalone Large Language Models (LLMs) by perceiving environments, making decisions, and executing actions. However, LLM agents still face challenges in tasks that require multiple decision-making steps. Estimating the value of actions in specific tasks is difficult when intermediate actions are neither appropriately rewarded nor penalized. In this paper, we propose leveraging a task-relevant Q-value model to guide action selection. Specifically, we first collect decision-making trajectories annotated with step-level Q values via Monte Carlo Tree Search (MCTS) and construct preference data. We then use another LLM to fit these preferences through step-level Direct Policy Optimization (DPO), which serves as the Q-value model. During inference, at each decision-making step, LLM agents select the action with the highest Q value before interacting with the environment. We apply our method to various open-source and API-based LLM agents, demonstrating that Q-value models significantly improve their performance. Notably, the performance of the agent built with Phi-3-mini-4k-instruct improved by 103% on WebShop and 75% on HotPotQA when enhanced with Q-value models, even surpassing GPT-4o-mini. Additionally, Q-value models offer several advantages, such as generalization to different LLM agents and seamless integration with existing prompting strategies.

Class-Incremental Learning (CIL) enables models to continuously learn new classes while mitigating catastrophic forgetting. Recently, Pre-Trained Models (PTMs) have greatly enhanced CIL performance, even when fine-tuning is limited to the first task. This advantage is particularly beneficial for CIL methods that freeze the feature extractor after first-task fine-tuning, such as analytic learning-based approaches using a least squares solution-based classification head to acquire knowledge recursively. In this work, we revisit the analytical learning approach combined with PTMs and identify its limitations in adapting to new classes, leading to sub-optimal performance. To address this, we propose the Momentum-based Analytical Learning (MoAL) approach. MoAL achieves robust knowledge memorization via an analytical classification head and improves adaptivity to new classes through momentum-based adapter weight interpolation, leading to forgetting outdated knowledge. Importantly, we introduce a knowledge rumination mechanism that leverages refined adaptivity, allowing the model to revisit and reinforce old knowledge, thereby improving performance on old classes. MoAL facilitates the acquisition of new knowledge and consolidates old knowledge, achieving a win-win outcome between plasticity and stability. Extensive experiments on various incremental settings show MoAL’s state-of-the-art performance1.

Multi-goal long-horizon problems are prevalent in real-world applications. The additional goal space introduced by multi-goal problems intensifies the spatial complexity of exploration; meanwhile, the long interaction sequences in long-horizon problems exacerbate the temporal complexity of exploration. Addressing the great exploration challenge posed by multi-goal long-horizon problems depends not only on the design of algorithms but also on the design of environments and the availability of demonstrations to assist in training. To facilitate the above research, we propose a multi-goal long-horizon Reinforcement Learning (RL) environment based on realistic fixed-wing UAV's velocity vector control, named VVC-Gym, and generate multiple demonstration sets of various quality. Through experimentation, we analyze the impact of different environment designs on training, assess the quantity and quality of demonstrations and their influence on training, and assess the effectiveness of various RL algorithms, providing baselines on VVC-Gym and its corresponding demonstrations. The results suggest that VVC-Gym is suitable for studying: (1) the influence of environment designs on addressing multi-goal long-horizon problems with RL. (2) the assistance that demonstrations can provide in overcoming the exploration challenges of multi-goal long-horizon problems. (3) the RL algorithm designs with the least possible impact from environment designs on the efficiency and effectiveness of training.

This paper addresses the challenge of discontinuity in goal-achievement capabilities observed in Goal-conditioned Reinforcement Learning (GCRL) algorithms. Through a theoretical analysis, we identify that the reuse of successful trajectories or policies during training can aid in achieving adjacent goals of achievable goals. However, the policy discrepancy between achievable and adjacent goals must be carefully managed to avoid both overly trivial and excessively large differences, which can respectively hinder policy performance. To tackle this issue, we propose a margin-based policy self-regularization approach that optimizes the policy discrepancies between adjacent desired goals to a minimal acceptable threshold. This method can be integrated into popular GCRL algorithms, such as GC-SAC, HER, and GC-PPO. Systematic evaluations across two robotic arm control tasks and a complex fixed-wing aircraft control task demonstrate that our approach significantly improves the continuity of goal-achievement abilities of GCRL algorithms, thereby enhancing their overall performance.

This paper addresses the challenge of Velocity Vector Control (VVC) for fixed-wing UAVs using Reinforcement Learning (RL) in the presence of imperfect demonstrations. The multi-objective and long-horizon nature of VVC introduces significant spatial and temporal complexities, complicating RL's exploration. While demonstration-based RL methods can help mitigate exploration challenges, their effectiveness is often limited by the quality of the provided demonstrations. To tackle this, we propose V-Pilot, a novel approach that integrates: (1) a controller equipped with a control law model to reduce action oscillation, thus alleviating temporal exploration issues, and (2) a VVC-specific training workflow for iterative policy refinement and demonstration quality improvement. This framework is designed to enhance the performance of demonstration-based RL under imperfect demonstrations. We evaluate V-Pilot on the fixed-wing UAV RL environment, VVCGym. Experimental results demonstrate that V-Pilot outperforms PID and Behavioral Cloning across multiple performance metrics.

Large language models (LLMs) enable versatile applications, with fine-tuning methods like LoRA optimizing them for specific domains. However, ensuring QoS for dynamic LoRA inference on heterogeneous devices is challenging due to memory contention from the shared LLM base model and KV cache. Existing solutions lack coordination between the KV cache and LoRA scheduling. We propose JointSerLoRA, a two-level inference system featuring: (1) a disaggregated architecture for KV cache consistency, (2) a global scheduler balancing adaptercache reuse and GPU load, and (3) a local scheduler with adaptive batching for prefix and adapter-aware grouping. Evaluations show that JointSerLoRA reduces end-to-end latency by 29 % and TTFT by$3.8 \times$compared to state-of-the-art systems.

The advent of cloud computing offers a novel mode of managing computing resources, enabling users to flexibly utilize required computing and storage resources through the network. Simultaneously, it allows large-scale computing centers and data centers to more effectively utilize their computing resources. With the rapid development of cloud computing technology, the exponential growth of massive data storage needs from numerous users has brought challenges in storage cost, performance, reliability, and security. Conventional single-objective optimization approaches, which concentrate exclusively on a singular performance metric, are increasingly inadequate to address the intricate requirements of cloud storage systems. In contrast, multi-objective optimization methods can simultaneously optimize multiple aspects of concern to users or operators, providing more comprehensive solutions for cloud computing services. In this paper, we first introduce the basic concepts of multi-objective optimization problems. Subsequently, we present a comprehensive review of multi-objective optimization applications in cloud storage systems, including the setting of optimization objectives, algorithm selection, and comparison methods of experimental results. Finally, this paper summarizes and discusses the current implementation of multi-objective evolutionary algorithms in cloud storage optimization, and provides an outlook on future research directions.

Audio classification constitutes a critical task aimed at assigning meaningful labels to audio recordings. Despite commendable efforts in this field, prior endeavors have encountered notable challenges. Firstly, the majority of existing solutions operate under the assumption of a fixed vocabulary for classification, neglecting the crucial need for systems to adapt continuously to dynamic data streams. Secondly, these approaches often assume that input data is fully annotated, a presumption that frequently diverges from practical scenarios. In response to these challenges, we present a novel semi-supervised class-incremental learning framework that utilizes memory replay with unlabeled data and introduces temporal consistency regularization to better alleviate catastrophic forgetting. Additional, we employ Iterative Projection and Matching to select the most informative samples for storage, addressing the issues of reservoir sampling while maintaining a balanced buffer. We showcase the framework's efficacy through a series of comprehensive experiments conducted on both the ESC-50 and Google Speech Commands datasets. The results demonstrate the remarkable performance of our proposed approach, particularly in few-shot scenarios.

Large language models (LLMs) have demonstrated great potential across diverse fields. However, their reasoning capabilities face challenges, especially when dealing with complex tasks. Existing work on enhancing LLMs' reasoning abilities mostly focuses on fields like mathematics and coding. Due to the specificity of domain data, these methods have poor generalization in specific areas such as medicine and materials. In this work, we propose an approach to enhance LLMs' reasoning capabilities in these fields lacking reasoning datasets. We construct reasoning datasets based on the logical relationships contained in their knowledge graphs. First, we clean the metadata, focus on a specific sub-field, and optimize data quality to construct a specialized knowledge graph. Then, we design seven knowledge graph patterns to extract instances, transform them into natural-language questions, and sample using the Deepseek-R1 model to build a reasoning dataset. We fine-tune the Qwen2.5-32B model with this dataset. Experimental results show that in the Relevance dimension, the score of Qwen2.5-32B improves from 45 to 98.57, with a remarkable increase of about 119.04%. In the Logic dimension, the score rises from 76.87 to 87.2, representing an increase of approximately 13.44%. This validates the effectiveness of our method. In these two aspects, the fine-tuned model reaches or approaches the level of advanced LLMs such as Deepseek-R1 and OpenAI 01, indicating that our approach can effectively enhance LLMs' reasoning ability.

Self-improving alignment leveraging large language models (LLMs) to automatically generate synthetic preference data has garnered significant attention as a means of reducing reliance on human labelers. These methods typically employ the LLM-as-a-judge mechanism, where the LLM generates responses and then employs itself to judge which response best aligns with the given prompt for curating the binary self-preferred dataset. However, these methods encounter two major challenges: (1) LLM-as-a-judge often produces error-prone evaluations, resulting in low-quality preference annotation, and (2) their optimization strategies often overlook the strength of preferences within binary pairs, leading to overfitting. This paper proposes a novel method, Preference-Strength-aware Optimization (PSO), to address these issues. Specifically, PSO frames the preference annotation process as a judgment token prediction task using the generative preference model to produce reliable judgments. The predicted judgment token indicates the preferred response and its corresponding probability reflects the disparity between responses, referred to as preference strength. Based on this strength, we introduce a new preference-strength-aware loss to adaptively reweight the impact of different response pairs on optimization, concentrating the model's learning on high-quality response pairs. Our experiments demonstrate that PSO significantly improves performance in preference benchmarks, achieving stronger alignment with human preferences, reducing verbose responses, and mitigating overfitting. Furthermore, PSO exhibits robust generalization and sample efficiency, offering a scalable and promising solution for LLM alignment without relying on human-annotated preferences.

Large Language Model (LLM) agents frequently produce sub-optimal actions when tackling complex, multi-step decision-making tasks. Employing self-critique to identify flaws and suggest enhancements is an effective strategy for refining actions. Although trajectory-level critique is commonly employed, it often fails to identify flawed steps accurately. In this paper, we introduce SLSC-MCTS, a method that integrates Monte Carlo Tree Search with Step-Level Self-Critique to enhance LLM agents during both testing and self-training phases. During decision tree expansion with SLSC-MCTS, the LLM agent initially generates an action, receives environmental feedback, and subsequently generates further actions via self-critique and refinement. Through multiple episodes of SLSC-MCTS, LLM agents can effectively utilize step-level critiques while disregarding ineffective ones based on node values, thereby incorporating the critiques more robustly. Additionally, our method further empowers LLM agents in a self-training manner, collecting training data from the constructed decision tree to iteratively fine-tune the LLM agents. The self-training data gathered via SLSC-MCTS is diverse and high-quality, which further enhances the reasoning, critiquing, and refining abilities of LLM agents. Experimental results demonstrate that SLSC-MCTS significantly improves LLM agents during testing, surpassing state-of-the-art baselines and achieving shorter task completion trajectories across information retrieval benchmarks such as WebShop and HotPotQA. After three iterations of self-training, LLM agents established by Llama-3.1-8B-Instruct show substantial improvement, even surpassing human experts in WebShop.

In the realm of open source software (OSS) development, the resolution of issues is not just a technical task but a pivotal activity that drives ongoing enhancement and secures a project's sustainability. Contributing to issues is the majority form for external contributors to take part in OSS projects. Although the significance of external contributors is recognized, their contributions in issue process still lacks full quantification and clarity, and the correlation between their contributions and the project productivity remains unclear. In response, we propose issue entropy, a novel metric that quantifies the complexity of event sequence in issue process applied to study the external contributions. The metric applies principles of information theory to scrutinize granular details within a project's issue-related activities, providing a unique lens to understand and assess external contributions. To explore the correlation between external contributions and project productivity, we employed issue entropy as a novel way to examine external contributions, and analyzed its correlation with new bugs, commits, and bug fix time, serving as proxies for OSS project productivity. Our findings reveal a strong positive correlation with new bugs, variable relationship with commit volume based on company sponsorship, and significant negative correlation with bug fix time. We also observe the significant interactions between external contributions and other factors, such as project age and the number of files. Moreover, issue entropy offers a new perspective on the health of the OSS project ecosystem, potentially supporting further research and practical applications.

Multi-view clustering (MVC) is a fast-growing research direction. However, most existing MVC works focus on concrete objects (e.g., cats, desks) but ignore abstract objects (e.g., knowledge, thoughts), which are also important parts of our daily lives and more correlated to cognition. Relational knowledge, as a typical abstract concept, describes the relationship between entities. For example, "Cats like eating fishes," as relational knowledge, reveals the relationship "eating" between "cats" and "fishes." To fill this gap, we first point out that MVC on relational knowledge is considered an important scenario. Then, we construct 8 new datasets to lay research grounds for them. Moreover, a simple yet effective relational knowledge MVC paradigm (RK-MVC) is proposed by compensating the omitted sample-global correlations from the structural knowledge information. Concretely, the basic consensus features are first learned via adopted MVC backbones, and sample-global correlations are generated in both coarse-grained and fine-grained manners. In particular, the sample-global correlation learning module can be easily extended to various MVC backbones. Finally, both basic consensus features and sample-global correlation features are weighted fused as the target consensus feature. We adopt 9 typical MVC backbones in this paper for comparison from 7 aspects, demonstrating the promising capacity of our RK-MVC.

Model-based offline reinforcement learning (RL) has made remarkable progress, offering a promising avenue for improving generalization with synthetic model rollouts. Existing works primarily focus on incorporating pessimism for policy optimization, usually via constructing a Pessimistic Markov Decision Process (P-MDP). However, the P-MDP discourages the policies from learning in out-of-distribution (OOD) regions beyond the support of offline datasets, which can under-utilize the generalization ability of dynamics models. In contrast, we propose constructing an Optimistic MDP (O-MDP). We initially observed the potential benefits of optimism brought by encouraging more OOD rollouts. Motivated by this observation, we present ORPO, a simple yet effective model-based offline RL framework. ORPO generates Optimistic model Rollouts for Pessimistic offline policy Optimization. Specifically, we train an optimistic rollout policy in the O-MDP to sample more OOD model rollouts. Then we relabel the sampled state-action pairs with penalized rewards, and optimize the output policy in the P-MDP. Theoretically, we demonstrate that the performance of policies trained with ORPO can be lower-bounded in linear MDPs. Experimental results show that our framework significantly outperforms P-MDP baselines by a margin of 30%, achieving state-of-the-art performance on the widely-used benchmark. Moreover, ORPO exhibits notable advantages in problems that require generalization.

Deep supervised learning has demonstrated strong capabilities; however, such progress relies on massive and expensive data annotation. Active Learning (AL) has been introduced to selectively annotate samples, thus reducing the human labeling effort. Previous AL research has focused on employing recently trained models to design sampling strategies, based on uncertainty or representativeness. Drawing inspiration from the issue of model forgetting, we propose a novel AL framework called Temporal Inconsistency-Based Active Learning (TIR-AL). In this framework, multiple snapshots of the models across consecutive cycles are jointly utilized to select samples with higher temporal inconsistency, by computing the proposed self-weighted nuclear norm metric. Furthermore, we introduce a consistency regularization term to mitigate the issue of forgetting. Together, these components make full use of the potential of data and facilitate effective interaction within the AL loop. To demonstrate the efficacy of TIR-AL, we conducted a set of experiments illustrating how our approach outperforms state-of-the-art methods without incurring any additional training costs.

Imitation learning heavily relies on the quality of provided demonstrations. In scenarios where demonstrations are imperfect and rare, a prevalent approach for refining policies is through online fine-tuning with reinforcement learning, in which a Kullback–Leibler (KL) regularization is often employed to stabilize the learning process. However, our investigation reveals that on the one hand, imperfect demonstrations can bias the online learning process, the KL regularization will further constrain the improvement of online policy exploration. To address the above issues, we propose Iterative Regularized Policy Optimization (IRPO), a framework that involves iterative offline imitation learning and online reinforcement exploration. Specifically, the policy learned online is used to serve as the demonstrator for successive learning iterations, with a demonstration boosting to consistently enhance the quality of demonstrations. Experimental validations conducted across widely used benchmarks and a novel fixed-wing UAV control task consistently demonstrate the effectiveness of IRPO in improving both the demonstration quality and the policy performance. Our code is available at https://github.com/GongXudong/IRPO.

Continuous integration, deployment and delivery (CI/CD) have become cornerstones of DevOps practices. In recent years, GitHub Action (GHA) has rapidly replaced the traditional CI/CD tools on GitHub, providing efficiently automated workflows for developers. With the widespread use and influence of GHA, it is critical to understand the existing problems that GHA developers face in their practices as well as the potential solutions to these problems. Unfortunately, we currently have relatively little knowledge in this area. To fill this gap, we conduct a large-scale empirical study of 6,590 Stack Overflow (SO) questions and 315 GitHub issues. Our study leads to the first comprehensive taxonomy of problems related to GHA, covering 4 categories and 16 sub-categories. Then, we analyze the popularity and difficulty of problem categories and their correlations. Further, we summarize 56 solution strategies for different GHA problems. We also distill practical implications of our findings from the perspective of different audiences. We believe that our study contributes to the research of emerging GHA practices and guides the future support of tools and technologies.

LLM-based workflows utilize Large Language Models (LLMs) for dynamic user requests, combining task planning and multiple machine learning (ML) models. Existing ML workflow platforms assume static structure deployment, neglecting dynamic orchestration for comprehensive workflows that fulfill users’ diverse requirements. Model selection in dynamic workflows involves various models and parallelism configurations, each with unique accuracy and efficiency trade-offs.To address these limitations, we introduce JointFlow, a solution offering LLM-based workflows as a service by dynamically constructing workflows across heterogeneous JointCloud infrastructures. It models and optimizes dynamic workflows, focusing on accuracy and efficiency trade-offs in model selection and parallelism configurations. A super-DAG represents the dynamic sub-task workflows from LLMs, profiling configurations across infrastructures. JointFlow also seeks the optimal workflow placement strategy. Experiments show it reduces serving costs while achieving throughput objectives compared to state-of-the-art methods.

Big data workflows are widely used in IoT, recommended systems, and real-time vision applications, and they continue to grow in complexity. These hybrid workflows consist of both resource-intensive batch jobs and latency-sensitive stream jobs. Examples include the data analytics workflow, which incorporates batch data transformations and low-latency querying, and the machine learning workflow, which processes stream data feature extraction before performing batch training and low-latency inference. However, existing research on workflow scheduling primarily focuses on either stream or batch workflows, neglecting the efficient scheduling of hybrid workflows that respect their diverse resource requirements and the costly data transfers between them.In this article, we propose a hybrid workflow model that defines the optimal placement of hybrid workflows (OHWP) as a bi-objective optimization problem. Our proposed model takes into account parameters related to inter-communication between stream and batch jobs, as well as the heterogeneous resources in JointCloud environment. Additionally, we present OHWP-PS (OHWP on a Pruned Space), a scheduling algorithm for hybrid workflows that minimizes both cost and latency by improving the initial population and dynamically updating the search space. The results demonstrate that the proposed OHWP-PS algorithm is effective and competitive across all experiments.

With the development of AI model ecology pioneered by the large language models, data plays a more important and diversified role than the traditional one. In recent years, data problems have been increasingly noticed and voiced by developers, and it is crucial to understand the existing challenges that they are facing in reality, as well as the potential features of these problems. Unfortunately, we currently have relatively little knowledge in this area. To fill this gap, we conduct an empirical study based on 2,154 posts and 785 issues by collecting model data problems from three online communities (i.e., GitHub, Stack Overflow, and Hugging Face). We present the first taxonomy of data problems around model production encountered by developers, covering 11 topics in 4 categories. Then, we analyze the popularity and the difficulty of the posts and issues in the derived topics. We distill many findings from the study and present the practical implications from the perspectives of different audiences.

With the rapid growth of large language models, cloud computing has become an indispensable component of the AI industry. Cloud service providers(CSPs) are establishing AI data centers to service AI workloads. In the face of this surging need for AI computing power, building a connected computing environment across various clouds and forming a JointCloud presents an attractive solution. However, scheduling AI tasks across multiple AI data centers within a JointCloud environment presents a significant challenge: how to balance users’ demands while ensuring CSPs’ fairness in scheduling. Existing research primarily focuses on optimizing scheduling quality with limited consideration for fairness. Therefore, this paper proposes a Fairness-Aware AI-Workloads Allocation method (F3A), a fair cross-cloud allocation technique for AI tasks. F3A utilizes Point and Token to reflect both the resource status and historical task allocations of AI data centers, enabling the consideration of users’ multidimensional demands and facilitating fair task allocation across multiple centers. In order to better assess the fairness of scheduling, we also devised a fairness indicator(FI), based on the Gini coefficient to measure the fairness of task allocation. The experimental results demonstrate that F3A consistently maintains FI within 0.1 across various cluster sizes and different task quantities, representing an improvement of 76.45% compared to classical fair scheduling algorithms round-robin. F3A exhibits commendable performance in ensuring fairness in task allocation while also demonstrating effectiveness in cost reduction and enhancing user satisfaction.

Active learning (AL) aims to select highly informative data points from an unlabeled dataset for annotation, mitigating the need for extensive human labeling effort. However, classical AL methods heavily rely on human expertise to design the sampling strategy, inducing limited scalability and generalizability. Many efforts have sought to address this limitation by directly connecting sample selection with model performance improvement, typically through influence function. Nevertheless, these approaches often ignore the dynamic nature of model behavior during training optimization, despite empirical evidence highlights the importance of dynamic influence to track the sample contribution. This oversight can lead to suboptimal selection, hindering the generalizability of model. In this study, we explore the dynamic influence based data selection strategy by tracing the impact of unlabeled instances on model performance throughout the training process. Our theoretical analyses suggest that selecting samples with higher projected gradients along the accumulated optimization direction at each checkpoint leads to improved performance. Furthermore, to capture a wider range of training dynamics without incurring excessive computational or memory costs, we introduce an additional dynamic loss term designed to encapsulate more generalized training progress information. These insights are integrated into a universal and task-agnostic AL framework termed Dynamic Influence Scoring for Active Learning (DISAL). Comprehensive experiments across various tasks have demonstrated that DISAL significantly surpasses existing state-of-the-art AL methods, demonstrating its ability to facilitate more efficient and effective learning in different domains.

Multimodal Large Language Models (MLLMs), by expanding the model's capabilities to perceive and interact through multi-modalities, have significantly enhanced performance across various tasks. The perception of vision is an important modality developed into LLM, enabling research in vision-language to continuously lead the cutting-edge advancements in the MLLM community. However, the standard pre-training pipeline on image-text pairs results in a limited model understanding of relationships between multiple images and texts, as well as visual details. Additionally, the setting of fine-tuning with a frozen visual backbone hinders the enhancement of visual representations on new data. These two issues lead to suboptimal performance in models for demonstrative instruction following about multiple images. This work introduces a novel framework called MLoEM, which first converts long multimodal data into an interleaved image-instruction format, and then adopts a fully autoregressive architecture model, allowing for more robust and coherent learning from naturally occurring multimodal documents than pair-based pipeline. Additionally, we incorporate the Low-Rank Adaptation (LoRA) fine-tuning method, enhancing visual representations while maintaining the stability of previously learned knowledge. Finally, we utilize ensemble methods to enhance model performance on tasks. To alleviate the storage overhead issue of parallel ensembles with large models, we design an ensemble approach that shares the MLLM while only switching the LoRA matrices. In the experiments, the proposed MLoEM shows superior performance on the testing set.

Continual learning (CL) empowers pre-trained vision-language (VL) models to efficiently adapt to a sequence of downstream tasks. However, these models often encounter challenges in retaining previously acquired skills due to parameter shifts and limited access to historical data. In response, recent efforts focus on devising specific frameworks and various replay strategies, striving for a typical learning-forgetting trade-off. Surprisingly, both our empirical research and theoretical analysis demonstrate that the stability of the model in consecutive zero-shot predictions serves as a reliable indicator of its anti-forgetting capabilities for previously learned tasks. Motivated by these insights, we develop a novel replay-free CL method named ZAF (Zero-shot Antidote to Forgetting), which preserves acquired knowledge through a zero-shot stability regularization applied to wild data in a plug-and-play manner. To enhance efficiency in adapting to new tasks and seamlessly access historical models, we introduce a parameter-efficient EMA-LoRA neural architecture based on the Exponential Moving Average (EMA). ZAF utilizes new data for low-rank adaptation (LoRA), complemented by a zero-shot antidote on wild data, effectively decoupling learning from forgetting. Our extensive experiments demonstrate ZAF's superior performance and robustness in pre-trained models across various continual VL concept learning tasks, achieving leads of up to 3.70\%, 4.82\%, and 4.38\%, along with at least a 10x acceleration in training speed on three benchmarks, respectively. Additionally, our zero-shot antidote significantly reduces forgetting in existing models by at least 6.37\%. Our code is available at https://github.com/Zi-Jian-Gao/Stabilizing-Zero-Shot-Prediction-ZAF.

Existing Goal-Conditioned Reinforcement Learning (GCRL) algorithms are built upon Hindsight Experience Replay (HER), which densifies rewards through hindsight replay and leverages historical goal-achieving information to construct a learning curriculum. However, when the task is characterized by a non-Markovian reward (NMR), whose computation depends on multiple steps of states and actions, HER can no longer densify rewards by treating a single encountered state as the hindsight goal. The lack of informative rewards hinders policy learning, resulting in rolling out failed trajectories. Consequently, the replay buffer is overwhelmed with failed trajectories, impeding the establishment of an applicable curriculum. To circumvent these limitations, we deviate from existing HER-based methods and propose an on-policy GCRL framework, GCPO, which is applicable to both multi-goal Markovian reward (MR) and NMR problems. GCPO consists of (1) Pre-training from Demonstrations, which pre-trains the policy to possess an initial goal-achieving capability, thereby diminishing the difficulty of subsequent online learning. (2) Online Self-Curriculum Learning, which first estimates the policy's goal-achieving capability based on historical evaluation information and then selects progressively challenging goals for learning based on its current capability. We evaluate GCPO on a challenging multi-goal long-horizon task: fixed-wing UAV velocity vector control. Experimental results demonstrate that GCPO is capable of effectively addressing both multi-goal MR and NMR problems.

Audio classification is a challenging task that requires categorizing audio data based on its content or characteristics. Existing approaches for audio classification rely either on supervised learning or fine-tuning based on self-supervised learning, both of which require manually labeled data. However, manually labeling audio datasets is a time-consuming and expensive process that limits the dataset's size. Moreover, the diversity of sound categories and class imbalances can further impede classification performance. To overcome these challenges, researchers have proposed various audio data augmentation methods. However, most of these methods focus less on augmentations combination and design and rely solely on waveform-based or spectrogram-based approaches. This paper presents an Automated Audio Augmentation (AAA) method for audio classification, which generates learnable and composable augmentation policies suitable for the audio classification task and can be employed in a plug-and-play manner. This method leverages both waveform-level and spectrogram-level augmentation, and a Bayesian optimization algorithm is proposed to search for composed augmentation policies. To the best of our knowledge, this is the first attempt to propose an automatic data augmentation method for audio classification tasks. Through large-scale empirical studies, we demonstrate that the proposed method outperforms previous competitive methods by a significant margin. We improve the average performance of multiple datasets by 6.421% and by 7.330% on few-shot scenarios, respectively.

Deep reinforcement learning has achieved encouraging performance in many realms. However, one of its primary challenges is the sparsity of extrinsic rewards, which is still far from solved. Complementary learning system theory suggests that effective human learning relies on two complementary learning systems utilizing short-term and long-term memories. Inspired by the fact that humans evaluate curiosity by comparing current observations with historical information, we propose a novel intrinsic reward, namely CLS-IR, which aims to address the problems caused by sparse extrinsic rewards. Specifically, we train a self-supervised predictive model with short-term and long-term memories via exponential moving averages. We employ the information gain between the two memories as the intrinsic reward, which does not incur additional training costs but leads to better exploration. To investigate the effectiveness of CLS-IR, we conduct extensive experimental evaluations; the results demonstrate that CLS-IR can achieve state-of-the-art performance on Atari games and DeepMind Control Suite.

The use of graph attention networks (GAT) in communication-enhanced multi-agent reinforcement learning (Comm-MARL) has become prevalent. While successful, GAT can lead to homogeneity in the strategies of message aggregation, which can severely limit multi-agent coordination. To address this challenge, we study the adjacency tensor of the communication graph. Then we define a new nuclear tensor rank and its convex surrogate, the normalized tensor nuclear norm to measure the homogeneity of message aggregation. Leveraging the norm, we further propose a plug-and-play regularizer on the adjacency tensor, named Normalized Tensor Nuclear Norm Regularization (NTNNR), to actively enrich the diversity of message aggregation during the training stage. NTNNR is agnostic to specific Comm-MARL algorithms and can be flexibly integrated with different graph-attention methods. Empirical results demonstrate that aggregating messages using NTNNR-enhanced GAT can improve the efficiency of the training and achieve higher asymptotic performance than existing message aggregation methods.

Federated learning (FL) enables distributed participants to collectively learn a strong global model without sacrificing their individual data privacy. Mainstream FL approaches require each participant to share a common network architecture and further assume that data are sampled IID across participants. However, in real-world deployments, participants may require heterogeneous network architectures; and the data distribution is almost non-uniform. To address these issues we introduce FedH2L, which is agnostic to the model architecture and robust to different data distributions across participants. In contrast to approaches sharing parameters or gradients, FedH2L relies on mutual distillation, exchanging only posteriors on a shared seed set between participants in a decentralized manner. This makes it extremely bandwidth efficient, model agnostic, and crucially produces models capable of performing well on the whole data distribution when learning from heterogeneous silos.

As a new stage in the development of the cloud computing paradigm, serverless computing has the high-level abstraction characteristic of shielding underlying details. This makes it extremely challenging for users to choose a suitable serverless platform. To address this, targeting the jointcloud computing scenario of heterogeneous serverless platforms across multiple clouds, this paper presents a jointcloud collaborative mechanism called FCloudless with cross-cloud detection of the full lifecycle performance of serverless platforms. Based on the benchmark metrics set that probe performance critical stages of the full lifecycle, this paper proposes a performance optimization algorithm based on detected performance data that takes into account all key stages that affect the performance during the lifecycle of a function and predicts the overall performance by combining the scores of local stages and dynamic weights. We evaluate FCloudless on AWS, AliYun, and Azure. The experimental results show that FCloudless can detect the underlying performance of serverless platforms hidden in the black box and its optimization algorithm can select the optimal scheduling strategy for various applications in a jointcloud environment. FCloudless reduces the runtime by 23.3% and 24.7% for cold and warm invocations respectively under cost constraints.

Visual Text Question Answering (VTQA) is a challenging task that requires answering questions pertaining to visual content by combining image understanding and language comprehension. The main objective is to develop models that can accurately provide relevant answers based on complementary information from both images and text, as well as the semantic meaning of the question. Despite ongoing efforts, the VTQA task presents several challenges, including multimedia alignment, multi-step cross-media reasoning, and handling open-ended questions. This paper introduces a novel generative framework called VTQAGen, which leverages a Multi- modal Attention Layer to combine image-text pairs and question inputs, as well as a BART-based model for reasoning and entity extraction from both images and text. The framework incorporates a step-based ensemble method to enhance model performance and generalization ability. VTQAGen utilizes an encoder-decoder generative model based on BART. Faster R-CNN is employed to extract visual regions of interest, while BART's encoder is modified to handle multi-modal interaction. The decoder stage utilizes the shift-predict approach and introduces step-based logits fusion to improve stability and accuracy. In the experiments, the proposed VTQAGen demonstrates superior performance on the testing set, securing second place in the ACM Multimedia Visual Text Question Answer Challenge.

Semantic networks, exemplified by the knowledge graph, serve as a means to represent knowledge by leveraging the structure of a graph. While the knowledge graph exhibits promising potential in the field of natural language processing, it suffers from incompleteness. This article focuses on the task of completing knowledge graphs by predicting linkages between entities, which is fundamental yet critical. Traditional methods based on translational distance struggle when dealing with unseen entities. In contrast, semantic matching presents itself as a potential solution due to its ability to handle such cases. However, semantic matching-based approaches necessitate large-scale datasets for effective training, which are typically unavailable in practical scenarios, hindering their competitive performance. To address this challenge, we propose a novel architecture for knowledge graphs known as LP-BERT, which incorporates a language model. LP-BERT consists of two primary stages: multi-task pre-training and knowledge graph fine-tuning. During the pre-training phase, the model acquires relationship information from triples by predicting either entities or relations through three distinct tasks. In the fine-tuning phase, we introduce a batch-based triple-style negative sampling technique inspired by contrastive learning. This method significantly increases the proportion of negative sampling while maintaining a nearly unchanged training time. Furthermore, we propose a novel data augmentation approach that leverages the inverse relationship of triples to enhance both the performance and robustness of the model. To demonstrate the effectiveness of our proposed framework, we conduct extensive experiments on three widely used knowledge graph datasets: WN18RR, FB15k-237, and UMLS. The experimental results showcase the superiority of our methods, with LP-BERT achieving state-of-the-art performance on the WN18RR and FB15k-237 datasets.

Sketches are widely deployed to represent network flows to support complex flow analysis. Typical sketches usually employ hash functions to map elements into a hash table or bit array. Such sketches still suffer from potential weaknesses upon throughput, flexibility, and functionality. To this end, we propose Ark filter, a novel sketch that stores the element information with either of two candidate buckets indexed by the quotient or remainder between the fingerprint and filter length. In this way, no further hash calculations are required for future queries or reallocations. We further extend the Ark filter to enable capacity elasticity and more functionalities (such as frequency estimation and top-$k$query). Comprehensive experiments demonstrate that, compared with Cuckoo filter, Ark filter has$2.08\times$,$1.34\times$, and$1.68\times$throughput of deletion, insertion, and hybrid query, respectively; compared with Quotient filter, Ark filter has$4.55\times$,$1.74\times$, and$22.12\times$throughput of deletion, insertion, and hybrid query, respectively; compared with Bloom filter, Ark filter has$2.55\times$and$2.11\times$throughput of insertion and hybrid query, respectively.

For most Open Source Software (OSS) projects, issues and Pull-requests (PR) are the primary means by which stakeholders of a project report and discuss software problems and code changes, and their descriptions are important for people to understand them. To help ensure the informational quality of issue/PR descriptions, GitHub introduced theissue/PR templatefeature, which pre-populates the description for anyone trying to open a new issue/PR. To better understand this feature, we report on a large-scale, mixed-methods empirical study of templates that explores contents, impacts, and perceptions. Our results show that templates typically contain elements to greet contributors, explain project guidelines, and collect relevant information. After template adoption, the monthly volume of incoming issues and PRs decreases, and issues have fewer monthly discussion comments and longer resolution duration. Although both contributors and maintainers positively rated the usefulness of templates from various aspects, they also reported challenges in using templates (e.g., excessive and irrelevant information request) and suggested potential improvements of the template feature (e.g., better user interaction and advanced automation). This work contributes to the informed use and targeted improvement of templates to enhance OSS practitioners’ collaboration and interaction.

While many forms of financial support are currently available, there are still many complaints about inadequate financing from software maintainers. In May 2019, GitHub, the world’s most active social coding platform, launched the Sponsor mechanism as a step toward more deeply integrating open source development and financial support. This paper collects data on 8,028 maintainers, 13,555 sponsors, and 22,515 sponsorships and conducts a comprehensive analysis. We explore the relationship between the Sponsor mechanism and developers along four dimensions using a combination of qualitative and quantitative analysis, examining why developers participate, how the mechanism affects developer activity, who obtains more sponsorships, and what mechanism flaws developers have encountered in the process of using it. We find a long-tail effect in the act of sponsorship, with most maintainers’ expectations remaining unmet, and sponsorship has only a short-term, slightly positive impact on development activity but is not sustainable. While sponsors participate in this mechanism mainly as a means of thanking the developers of OSS that they use, in practice, the social status of developers is the primary influence on the number of sponsorships. We find that both the Sponsor mechanism and open source donations have certain shortcomings and need further improvements to attract more participants.

We present an approach to learn voice-face representations from the talking face videos, without any identity labels. Previous works employ cross-modal instance discrimination tasks to establish the correlation of voice and face. These methods neglect the semantic content of different videos, introducing false-negative pairs as training noise. Furthermore, the positive pairs are constructed based on the natural correlation between audio clips and visual frames. However, this correlation might be weak or inaccurate in a large amount of real-world data, which leads to deviating positives into the contrastive paradigm. To address these issues, we propose the cross-modal prototype contrastive learning (CMPC), which takes advantage of contrastive methods and resists adverse effects of false negatives and deviate positives. On one hand, CMPC could learn the intra-class invariance by constructing semantic-wise positives via unsupervised clustering in different modalities. On the other hand, by comparing the similarities of cross-modal instances from that of cross-modal prototypes, we dynamically recalibrate the unlearnable instances' contribution to overall loss. Experiments show that the proposed approach outperforms state-of-the-art unsupervised methods on various voice-face association evaluation protocols. Additionally, in the low-shot supervision setting, our method also has a significant improvement compared to previous instance-wise contrastive learning.

As the new paradigm of JointCloud Computing matures, enterprises are trying to build multiple Kubernetes clusters on different clouds to deploy tasks, with the advantages of disaster backup, low latency, and avoidance of single vendor lock-in, etc. Tasks in a JointCloud environment, always have highly diversified resource demands on CPU, memory, disk, and network. However, the mismatch between these tasks and heterogeneous clusters can easily cause many resource fragments, resulting in low resource availability. Therefore, the task scheduling strategy is the key to solving the above problem. The existing task schedule strategies for multi-clusters are always aiming at clusters’ load balancing instead of increasing the resource availability. In this paper, we propose a dynamic task scheduling framework with the design of multi-cluster resource high-availability schedule strategy (MRASS) based on historical task resource consumption. MRASS conducts a cooperation model between multiple clusters and tasks, and proposes an indicator of resource availability, which is used to optimize the proportion of remaining resources of the cluster to keep approaching the proportion of resource requirements of future tasks, thereby execute more tasks within limited resources. Extensive numerical results confirm that the strategy has stable performance and performs well with different initial cluster resource setting, task resource type and task number. Compared with the existing algorithm, MRASS can place up to 20% more tasks, and the success rate of first placement of tasks can reach over 98%.

Docker building is a critical component of containerized workflow, which automates the process by which sources are packaged and transformed into container images. If not run properly, Docker builds can bring long durations (i.e., slow builds), which increases the cost in human and computing resources, and thus inevitably affect the software development. However, the current status and remedy for the duration cost in Docker builds remain unclear and need an in-depth study. To fill this gap, this paper provides the first empirical investigation on 171,439 Docker builds from 5,833 open source software (OSS) projects. Starting with an exploratory study, the Docker build durations can be characterized in real-world projects, and the developers’ perceptions of slow builds are obtained via a comprehensive survey. Driven by the results of our exploratory study, we propose a prediction modeling of Docker build duration, leveraging 27 handcrafted features from build-related context and configuration and 8 regression algorithms for the prediction task. Our results demonstrate that Random Forest model provides the superior performance with a Spearman’s correlation of 0.781, outperforming the baseline random model by 82.9% in RMSE, 90.6% in MAE, and 94.4% in MAPE, respectively. The implications of this study will facilitate research and assist practitioners in improving the Docker build process.

Background: The Pull-Request (PR) model is a widespread approach adopted by open source software (OSS) projects to support collaborative software development. However, it is often challenging to continuously evaluate and revise PRs in several iterations of code reviewsinvolving technical and social aspects. Aim: Our objective is twofold: identifying best practices for effective collaboration in continuous PR improvement and uncovering problems that deserve special attention to improve collaboration efficiency and productivity. Method: We conducted a mixed-methods empirical study of repeatedly revised PRs (i.e. those that have undergone a high number of revisions). Historical trace data of five long-lived popular GitHub projects were used for manual investigation of practices for requesting changes to PRs and reasons for nonacceptance of repeatedly revised PRs. Surveys of OSS practitioners were conducted to evaluate the results of manual analysis and to provide additional insights into developers' willingness regarding PR revisions and factors causing avoidable revisions in practice. Results: The main results of our research were as follows: (1) We identified 15 code review practices for requesting changes to PRs, among which practices with respect to explaining the reasoning behind requested changes and tracking the progress of PR review and revision were undervalued by reviewers; (2) While submitters can in general undergo 1-5 rounds of revisions, they are willing to offer more revisions when they are in a friendly community and receive helpful feedback; (3) We revealed 11 factors causing avoidable revisions regarding to reviewers' feedback, code review policy, pre-submission issues, and implementation of new revisions; and (4) Nonacceptance of repeatedly revised PRs was due mainly to inactivity of submitters or reviewers and being superseded for better maintenance. Finally, based on these findings, we proposed recommendations and implications for OSS practitioners and tool designers to facilitate efficient collaboration in PR revisions.

With the development of Blockchain technology, permissioned Blockchains are getting more and more attention from researchers because applications based on permissioned Blockchains are more practical and easier to be carried out. This paper aims to design a dedicated consensus protocol for permissioned Blockchains. The existing consensus protocols applied to permissioned Blockchains are either derived from public Blockchains such as Proof of Work (PoW) or Proof of Stake (PoS), with full decentralization, resulting in low transaction processing efficiency; or derived from traditional Byzantine fault-tolerant (BFT) consensus protocols such as Practical BFT (PBFT) or HoneyBadgerBFT, with high communication complexity of the consensus process, resulting in low scalability. Therefore, we propose a dedicated consensus protocol for permissioned Blockchains called Votes-as-a-Proof (VaaP) with high transaction processing efficiency while ensuring high scalability. Every node in VaaP runs a simple consensus process based on voting in parallel. Faulty nodes will only deprive themselves of using consensus service. We present the comparison of VaaP and Sphinx, one of the state-of-the-art consensus protocols, analytically and experimentally (up to 500 nodes). The results indicate that VaaP outperforms Sphinx in throughput, latency and scalability.

The great success of numerous community-based open source software (OSS) is based on volunteers continuously submitting contributions, but ensuring sustainability is a persistent challenge in OSS communities. Although the motivations behind and barriers to OSS contributors’ joining and retention have been extensively studied, the impacts of, reasons for and solutions to contribution abandonment at the individual level have not been well studied, especially for pull-based development. To bridge this gap, we present an empirical study on pull request abandonment based on a sizable dataset. We manually examine 321 abandoned pull requests on GitHub and then quantify the manual observations by surveying 710 OSS developers. We find that while the lack of integrators’ responsiveness and the lack of contributors’ time and interest remain the main reasons that deter contributors from participation, limitations during the processes of patch updating and consensus reaching can also cause abandonment. We also show the significant impacts of pull request abandonment on project management and maintenance. Moreover, we elucidate the strategies used by project integrators to cope with abandoned pull requests and highlight the need for a practical handover mechanism. We discuss the actionable suggestions and implications for OSS practitioners and tool builders, which can help to upgrade the infrastructure and optimize the mechanisms of OSS communities.

OSS projects are being developed by globally distributed contributors, who often collaborate through the pull-based model today. While this model lowers the barrier to entry for OSS developers by synthesizing, automating and optimizing the contribution process, coordination among an increasing number of contributors remains as a challenge due to the asynchronous and self-organized nature of distributed development. In particular, duplicate contributions, where multiple different contributors unintentionally submit duplicate pull requests to achieve the same goal, are an elusive problem that may waste effort in automated testing, code review and software maintenance. While the issue of duplicate pull requests has been highlighted, to what extent duplicate pull requests affect the development in OSS communities has not been well investigated. In this paper, we conduct a mixed-approach study to bridge this gap. Based on a comprehensive dataset constructed from 26 popular GitHub projects, we obtain the following findings: (a) Duplicate pull requests result in redundant human and computing resources, exerting a significant impact on the contribution and evaluation process. (b) Contributors’ inappropriate working patterns and the drawbacks of their collaborating environment might result in duplicate pull requests. (c) Compared to non-duplicate pull requests, duplicate pull requests have significantly different features, e.g., being submitted by inexperienced contributors, being fixing bugs, touching cold files, and solving tracked issues. (d) Integrators choosing between duplicate pull requests prefer to accept those with early submission time, accurate and high-quality implementation, broad coverage, test code, high maturity, deep discussion, and active response. Finally, actionable suggestions and implications are proposed for OSS practitioners.

To encourage developers' volunteer contributions, modern programming question and answer (Q&A) sites like Stack Overflow (SO) employ gamified incentive mechanisms such as reputation and badges. Understanding developers' motivations in the presence of gamification and the relationship between their motivations and behavioral outcomes is crucial for community building and designing good incentive mechanisms. Grounded on self-determination theory, we conducted a survey with 938 developers who participate in SO to understand their participation motivations and incentive perceptions. By connecting the survey responses with the SO data, we quantitatively analyzed how the developers' motivations and satisfaction of needs relate to their effort and contribution quality. Our main findings are as follows: (1) despite the presence of gamified incentive mechanisms, developers are mainly motivated by intrinsic motivation to participate in SO; (2) developers who have strong motivations to gain gamification rewards are associated with higher intrinsic and integrated motivations, while developers with more development experiences are less motivated by the gamified incentives; (3) both extrinsic motivations (in terms of career prospects) and intrinsic motivations (regarding self-improvement and helping others) can motivate developers to make high-quantity and high-quality contributions; and (4) high-level satisfaction of needs for competency and autonomy has a positive effect on developers making high-quantity and high-quality contributions and addressing difficult problems. Based on these findings, we discuss implications for developer motivation and gamification in the crowdsourcing context and for the mechanism design of gamified crowdsourced platforms.

B-mode ultrasound tongue imaging is widely used to visualize the tongue motion, due to its appearing properties. Extracting the tongue surface contour in the B-mode ultrasound image is still a challenge, while it is a prerequisite for further quantitative analysis. Recently, deep learning-based approach has been adopted in this task. However, the standard deep models fail to address faint contour when the ultrasound wave goes parallel to the tongue surface. To address the faint or missing contours in the sequence, we explore the shape consistency-based regularizer, which can take sequential information into account. By incorporating the regularizer, the deep model not only can extract frame-specific contours, but also can enforce the similarity between the contours extracted from adjacent frames. Extensive experiments are conducted both on the synthetic and real ultrasound tongue imaging dataset and the results demonstrate the effectiveness of proposed method. To better promote the research in this field, we have released our code at1.

Summary The blockchain is a distributed ledger that records all transactions and operations in a shared manner. Public blockchains such as Bitcoin realize decentralization at the cost of mining overhead, which is not suitable for real‐life scenarios requiring high throughput. Techniques such as the consortium blockchain improve efficiency through partial decentralization. However, the consensus algorithms used in the existing state‐of‐the‐art consortium blockchains face many challenges when dealing with commercial applications. For example, the high communication overhead hinders the scalability of PBFT‐based consensus algorithms even though they are efficient at small scale. Hashgraph, one of the most popular Directed Acyclic Graph‐based (DAG‐based) consensus algorithms, achieves good performance in scalability; however, it does not allow users' dynamic participation. To deal with these challenges, we propose Jointgraph, a Byzantine fault‐tolerance consensus algorithm for consortium blockchains based on DAG. In Jointgraph, transactions are packed into events and validated by no less than 2/3 of all members. A supervisor is introduced in our design, who monitors member behaviors and improves consensus efficiency. Simulation results demonstrate that Jointgraph outperforms Hashgraph in both throughput and latency.

Summary Sharing trusted data among trusted stakeholders is very important to large‐scale Internet of Things (IoT) applications. However, the entities and organizations involved in IoT naturally lack trusted relationships, which poses significant challenges to the above vision. Specifically, the first challenge is to ensure that the data in the physical world can be objectively and truly injected into the information world of IoT. The second is to ensure the credibility of the entities' identities in IoT. The third is to ensure the authenticity of data, the credibility of identity, and the reliable transmission of data when a third trusted party is unable to provide the expected trusted services. In view of the above challenges, this paper proposes a secure and lightweight triple‐trusting architecture (SLTA), which fully uses a blockchain‐related supporting technology. The architecture includes an oracle‐based data collection mechanism, which ensures that the data collected from edge devices of IoT cannot be modified, and the distributed identity management mechanism, which enhances personal privacy, security, and control of digital identities. Furthermore, a series of innovative designs for applying the blockchain to special large‐scale cooperation scenario in IoT are proposed, which is also a part of the key mechanisms of the SLTA. The innovative design includes a new software‐defined blockchain structure model and a lightweight Byzantine fault‐tolerant algorithm that provides credible support for decentralized data collection, identity management, and data transfer, as well as low‐overhead sequential storage mechanism.

JCLedger is a BlockChain-based distributed ledger for JointCloud that can improve the reliability and convenience of cloud resource exchanges by empowering cooperation among multiple clouds. The biggest challenge for the implementation of JCLedger is the approach to consensus. The existing consensus algorithms for the public BlockChain, such as proof of work (PoW) or proof of stake (PoS) does not apply to the JointCloud, because they require a massive computing power with a low throughput or monopoly risk. In this paper, we propose a practical Byzantine-fault-tolerance (PBFT)-based consensus algorithm called proof of previous transactions (PoPT), in which the accountants are selected by a specific hash function from a certain number of candidates. The candidates are chosen according to the users' participation in JointCloud, and only candidates that join the PBFT-based consensus process instead of all users. We also propose a new BlockChain structure for parallel accounting to improve the scalability of JCLedger, and a consistent hashing algorithm is used to assign the transactions to different accountants. Simulation experiments show that the PoPT can shield the unequal computing power of the users to provide them equal accounting opportunities, and the parallel accounting can handle the massive and high-frequency transactions in JointCloud more efficiently.

Code clone helps improving programming productivity, while at the same time leads to many negative effects on software maintenance. Many approaches have been proposed to detect clones, but most of them fail on detecting low similarity code snippets. In this paper, we propose a Siamese network which links two recursive autoencoders (RAE) with a comparator network for clone detection. The unweighted recursive autoencoder is designed to learn code representation and then the comparator network is employed for similarity evaluation. In this Siamese network, it takes full advantages of lexical, semantic and structure information, and achieves high accuracy in revealing tiny similarity. We conduct comprehensive experiments on BigCloneBench using tagged clones as well as the whole repository respectively. The results suggest that our approach achieves good accuracy, and its recall reaches 93.02 % in WT3/T4, which outperforms state-of-the-art.

Docker is one of the most popular containerization tools in current DevOps practice. Particularly, Dockerfile plays an important role in the Docker-based software development process by specifying the commands and build environment of Docker containers. As a project progresses through its development stages, the content of the Dockerfile may be revised many times. Previous studies have examined Dockerfile usage in open-source projects. However, little is known about the details of Dockerfile changes in practice. In this paper, we conduct an empirical study on Dockerfile changes for 4,110 open-source projects hosted on GitHub. Based on the Dockerfile data, we measure the frequency, magnitude, and instructions of Dockerfile changes and report how Dockerfile co-changed with other files. To explore the relationship between Dockerfile changes and project outcomes, i.e., popularity, success, and productivity, we also develop regression models, by controlling for various confounds. Our findings help to characterize and understand Dockerfile changes and motivate the need for collecting more empirical evidence.

Container technologies are being widely used in large scale production cloud environments, of which Docker has become the de-facto industry standard. In practice, Docker builds often break, and a large amount of efforts are put into troubleshooting broken builds. Prior studies have evaluated the rate at which builds in large organizations fail. However, there is still a lack of early warning methods for predicting the Docker build result before the build starts. This paper provides a first attempt to propose an automatic method named PDBR. It aims to use the configuration semantic features extracted by AST and the machine learning algorithms to predict build result in the cloud-based container environment. The evaluation experiments based on more than 36,000 collected Docker builds show that PDBR achieves 73.45%-91.92% in F1 and 29.72%-72.16% in AUC. We also demonstrate that different ML classifiers have significant and large effects on the PDBR AUC performance.

Computation offloading has been widely recognized as an effective way to promote the capabilities of resource-constrained mobile devices. The past years have seen a renewed importance of this technology in the emerging field of mobile robots. However, a significant feature of robots compared to traditional mobile computing devices (e.g. smartphones) is that they must collaborate to solve complex tasks in the physical world in many cases, which implies intensive data exchange among the robot peers. This characteristic has not been dealt with in-depth in traditional computation offloading research. In this paper, we propose an approach called Cooperative Offloading, which takes into account the cooperation among mobile devices as well as the communication it brings in computation offloading. Firstly, we propose the offloading decision approach that deals with offloading from a set of robots that cooperate to finish a specific task. Secondly, we present a set of mechanisms to optimize the data transfer path on network topology in multi-robot applications, which can significantly reduce the bandwidth consumption of the robot wireless network. Based on the cooperative offloading approach, we realized a cloud robotics framework called Cloudroid Swarm. Evaluations based on real-life applications have shown that Cloudroid Swarm brings more than five times performance promotion compared to the setup without offloading or with individual computation offloading.

The application of beauty and personal-care product retrieval seems to be evident in our daily life, and it has attracted increasing research interests during the last decade. However, the retrieval task is suffered from different image variations and complicated backgrounds. Recent works have demonstrated that Generalized-attention Regional Maximal Activation of Convolutions (GRMAC) descriptor can provide state-of-the-art performance for the retrieval task. However, GRMAC descriptor is restrained from the essentially limited property of the employed feature from a single layer. Features from a single layer are not robust enough for scale variations, shape deformation, and heavy occlusion. In this paper, we propose a novel descriptors, named Multi-Scale Generalized Attention-Based Regional Maximum Activation of Convolutions (MS-GRMAC). This method introduces multi-scale generalized attention mechanism to reduce the influence of scale variations, thus, can boost the performance of the retrieval task. To empirically investigate the effectiveness of the proposed approach, we conduct extensive experiments on the dataset containing more than half-million personal-care products (Perfect-500K) and obtain satisfactory results without ensemble.

Social media popularity estimation refers to predict the post's popularity using multimodal contents. The prediction performance heavily relies on the feature extraction part and fully leveraging multimodal heterogeneous data is of a great challenge in the practical settings. Despite remarkable progress have been made, most of the previous attempts are restrained from the essentially limited property of the employed single modality. Inspired by the recent success of multimodal learning, we propose a novel multimodal deep learning framework for the popularity prediction task, which aims to leverage the complementary knowledge from different modalities. Moreover, an attention mechanism is introduced in our framework, with the goal to assign large weights to specified modalities during the training and inference phases. To empirically investigate the effectiveness and robustness of the proposed approach, we conduct extensive experiments on the 2020 SMP challenge. The obtained results show that the proposed framework outperforms related approaches.

Docker containers have become the de-facto industry standard. Docker builds often break, and a large amount of efforts are put into troubleshooting broken builds. Prior studies have evaluated the rate at which builds in large organizations fail. However, little is known about the frequency and fix effort of failures that occur in Docker builds of open-source projects. This paper provides a first attempt to present a preliminary study on 857,086 Docker builds from 3,828 open-source projects hosted on GitHub. Using the Docker build data, we measure the frequency of broken builds and report their fix time. Furthermore, we explore the evolution of Docker build failures across time. Our findings help to characterize and understand Docker build failures and motivate the need for collecting more empirical evidence.

Off-Policy Actor-Critic (OffP-AC) methods have proven successful in a variety of continuous control tasks. Normally, the critic's action-value function is updated using temporal-difference, and the critic in turn provides a loss for the actor that trains it to take actions with higher expected return. In this paper, we introduce a flexible and augmented meta-critic that observes the learning process and meta-learns an additional loss for the actor that accelerates and improves actor-critic learning. Compared to existing meta-learning algorithms, meta-critic is rapidly learned online for a single task, rather than slowly over a family of tasks. Crucially, our meta-critic is designed for off-policy based learners, which currently provide state-of-the-art reinforcement learning sample efficiency. We demonstrate that online meta-critic learning benefits to a variety of continuous control tasks when combined with contemporary OffP-AC methods DDPG, TD3 and SAC.

Abstract Social coding site GitHub provides developers with many management tools to facilitate project maintenance and developer collaboration.Milestonetool, in particular, plays an important role in organizing and tracking progress on groups of issues or pull requests in a project. However, few research has analyzed the milestone tool, even though it has been used in practice for a long time. In this paper, we want to address this literature gap and present an ongoing work aimed at investigating the use of the milestone tool in GitHub open‐source projects. We conduct a mixed‐methods analysis in a large‐scale dataset of GitHub projects, to help developers gain some insights into the milestone tool, including its usage, benefits, and limitations. We quantitatively investigate the basic adoption of milestone tool and its correlation with project properties. We also survey developers to understand the reasons for using milestone tool or not and their perceptions of the milestone tool. We find that certain types of projects use milestone tool more than others. Adopting the milestone tool is associated with more commits, more releases, and more project popularity, but the current milestone tool also has some limitations. These observations can then be forwarded to the GitHub community for follow‐up and can result in them potentially making a better milestone tool.

High quality labeled datasets have allowed deep learning to achieve impressive results on many sound analysis tasks. Yet, it is labor-intensive to accurately annotate large amount of audio data, and the dataset may contain noisy labels in the practical settings. Meanwhile, the deep neural networks are susceptive to those incorrect labeled data because of their outstanding memorization ability. In this article, we present a novel framework, named CrossFilter, to combat the noisy labels problem for audio tagging. Multiple representations (such as, Logmel and MFCC) are used as the input of our framework for providing more complementary information of the audio. Then, though the cooperation and interaction of two neural networks, we divide the dataset into curated and noisy subsets by incrementally pick out the possibly correctly labeled data from the noisy data. Moreover, our approach leverages the multi-task learning on curated and noisy subsets with different loss function to fully utilize the entire dataset. The noisy-robust loss function is employed to alleviate the adverse effects of incorrect labels. On both the audio tagging datasets FSDKaggle2018 and FSDKaggle2019, empirical results demonstrate the performance improvement compared with other competing approaches. On FSDKaggle2018 dataset, our method achieves state-of-the-art performance and even surpasses the ensemble models.

B-mode ultrasound tongue imaging is widely used in the speech production field. However, efficient interpretation is in a great need for the tongue image sequences. Inspired by the recent success of unsupervised deep learning approach, we explore unsupervised convolutional network architecture for the feature extraction in the ultrasound tongue image, which can be helpful for the clinical linguist and phonetics. By quantitative comparison between different unsupervised feature extraction approaches, the denoising convolutional autoencoder (DCAE)-based method outperforms the other feature extraction methods on the reconstruction task and the 2010 silent speech interface challenge. A Word Error Rate of 6.17% is obtained with DCAE, compared to the state-of-the-art value of 6.45% using Discrete cosine transform as the feature extractor. Our codes are available at https://github.com/DeePBluE666/Source-code1.

A challenge in speech production research is to predict future tongue movements based on a short period of past tongue movements. This study tackles speaker-dependent tongue motion prediction problem in unlabeled ultrasound videos with convolutional long short-term memory (ConvLSTM) networks. The model has been tested on two different ultrasound corpora. ConvLSTM outperforms 3-dimensional convolutional neural network (3DCNN) in predicting the 9thframes based on 8 preceding frames, and also demonstrates good capacity to predict only the tongue contours in future frames. Further tests reveal that ConvLSTM can also learn to predict tongue movements in more distant frames beyond the immediately following frames. Our codes are available at: https://github.com/shuiliwanwu/ConvLstm-ultrasound-videos.

With growing scale of the data volume and neural network size, we have come into the era of distributed deep learning. High-performance training and inference on distributed computing systems has been attracting increasing research attention in both academia and industry. Meanwhile, diversity of existing machine learning frameworks (e.g. TensorFlow, Pytorch and MXNet) and the explosion of deep learning hardwares (e.g. CPUs, GPUs, FPGAs and ASICs) bring more challenges for users to leverage new deep learning technologies and accelerating capability of hardware devices. We firstly search around the state-of-the-art work in the area which open our mind to take a vision upon the future deep learning framework. Then, we propose HPDL, a general framework for high-performance distributed deep learning which is compatible with existing frameworks and adaptive to various hardware architectures. At last, we discuss and foresee the key technologies fulfilling high-performance and large-scale deep learning, including optimization algorithm, hybrid communication mechanism, model parallelization, resource scheduling and single-node execution optimization.

Sound event detection is intended to analyze and recognize the sound events in audio streams and it has widespread applications in real life. Recently, deep neural networks such as convolutional recurrent neural networks have shown state-of-the-art performance in this task. However, the previous methods were designed and implemented on devices with rich computing resources, and there are few applications on mobile devices. This paper focuses on the solution on the mobile platform for sound event detection. The architecture of the solution includes offline training and online detection. During offline training process, multi model-based distillation method is used to compress model to enable real-time detection. The online detection process includes acquisition of sensor data, processing of audio signals, and detecting and recording of sound events. Finally, we implement an application on the mobile device that can detect sound events in near real time.

Recently, deep learning has witnessed dramatic progress in the medical image analysis field. In the precise treatment of cancer immunotherapy, the quantitative analysis of PD-L1 immunohistochemistry is of great importance. It is quite common that pathologists manually quantify the cell nuclei. This process is very time-consuming and error-prone. In this paper, we describe the development of a platform for PD-L1 pathological image quantitative analysis using deep learning approaches. As point-level annotations can provide a rough estimate of the object locations and classifications, this platform adopts a point-level supervision model to classify, localize, and count the PD-L1 cells nuclei. Presently, this platform has achieved an accurate quantitative analysis of PD-L1 for two types of carcinoma, and it is deployed in one of the first-class hospitals in China.

In the GitHub platform, social media profile images are one of many visual components of developers. Besides, developer activities such as reporting issues or following other developers are regarded as important development and self-expression behaviors. However, to the best of our knowledge, no study has yet been conducted to study the relationship between GitHub developer activities and profile images. In this paper, we aim to investigate the relationship between developer activities and profile images to gain some insights into the developers' internal properties. During our experiments, we manually classify profile images into seven categories. Next, we investigate the relationship between developer's demographic information and developer activity. Further, using logistic regression analysis, when controlled for various variables, we statistically identify and quantify the relationships between developer activities and profile image categories. We find that several profile image categories significantly correlate with developer's demographic information and activities. We also provide a rich resource of research ideas for further study. Our examination and analysis provide insights into the developers' internal properties when using different profile images. Moreover, this study is the first step in understanding the relationship between developer activities and profile images on GitHub.

With the deep integration of software collaborative development and social networking, social coding represents a new style of software production and creation paradigm. Because of their good flexibility and openness, a large number of external contributors have been attracted to the open-source communities. They are playing a significant role in open-source development. However, the open-source development online is a globalized and distributed cooperative work. If left unsupervised, the contribution process may result in inefficiency. It takes contributors a lot of time to find suitable projects or tasks from thousands of open-source projects in the communities to work on. In this paper, we propose a new approach called “RepoLike,” to recommend repositories for developers based on linear combination and learning to rank. It uses the project popularity, technical dependencies among projects, and social connections among developers to measure the correlations between a developer and the given projects. Experimental results show that our approach can achieve over 25% of hit ratio when recommending 20 candidates, meaning that it can recommend closely correlated repositories to social developers.

Incomplete multi-view clustering optimally integrates a group of pre-specified incomplete views to improve clustering performance. Among various excellent solutions, multiple kernel $k$k-means with incomplete kernels forms a benchmark, which redefines the incomplete multi-view clustering as a joint optimization problem where the imputation and clustering are alternatively performed until convergence. However, the comparatively intensive computational and storage complexities preclude it from practical applications. To address these issues, we propose Late Fusion Incomplete Multi-view Clustering (LF-IMVC) which effectively and efficiently integrates the incomplete clustering matrices generated by incomplete views. Specifically, our algorithm jointly learns a consensus clustering matrix, imputes each incomplete base matrix, and optimizes the corresponding permutation matrices. We develop a three-step iterative algorithm to solve the resultant optimization problem with linear computational complexity and theoretically prove its convergence. Further, we conduct comprehensive experiments to study the proposed LF-IMVC in terms of clustering accuracy, running time, advantages of late fusion multi-view clustering, evolution of the learned consensus clustering matrix, parameter sensitivity and convergence. As indicated, our algorithm significantly and consistently outperforms some state-of-the-art algorithms with much less running time and memory.

Server consolidation is critical for energy efficiency of cloud-scale computing. In production environments like Aliyun, which is one of the largest public cloud platforms in the world, server consolidation has several challenges. First, the widespread use of local storage remarkably increases the migration cost (time). Second, the resource utilization of service instances varies over time, which may result in migration oscillation. Third, server consolidation must follow practical constraints. E.g., instances can only be migrated within a maintenance window, and both the resource utilization and the number of instances on a server are bounded. This paper designs and implements C4, a Cost-Efficient Consolidating Service for Aliyun's Cloud-Scale Computing, which Aliyun uses to consolidate servers. We analyze user pattern, resource utilization, and migration cost in Aliyun, showing that traditional utilization-based consolidation approaches cannot meet the needs in production environments, especially for the local-storage-based computing. This motivates us to propose the migration cost model, by which to select servers with the minimum migration time to release. We use the Worst-Fit heuristic to migrate instances to balance the load. Evaluation shows that C4 achieves cost-efficient, load-balanced, and oscillation-free consolidating service. We describe experience with over one year of C4 production deployment, lessons learned, and areas for future work.

Containerization is a software development approach aimed at packaging an application together with all its dependencies and execution environment in a light-weight, self-contained unit, of which Docker has become the de-facto industry standard. By defining the specific Docker image architecture and building orders, dockerfile plays an important role in the Docker-based containerization process. Understanding the evolution of dockerfile and which dockerfile architecture attributes enhance dockerfile quality and reduce image build latency can benefit the efficient processing of containerization. In this paper, we perform an empirical study on a large dataset of 2,840 projects to shed light on the impact of dockerfile evolutionary trajectories on quality and latency in the Docker-based containerization. Based on the six categories of dockerfile evolutionary trajectories we discovered, we build two regression models to explore the impact of dockerfile evolutionary trajectories and specific architecture attributes on dockerfile quality and image build latency, which derives a number of suggestions for practitioners.

Virtual machines (VMs) and processes are two important abstractions for cloud virtualization, where VMs usually install a complete operating system (OS) executing user processes. Although existing in different layers in the virtualization hierarchy, VMs and processes have overlapped functionalities. For example, they are both intended to provide execution abstraction (e.g., physical/virtual memory address space), and share similar objectives of isolation, cooperation and scheduling. However, neither of them could provide the benefits of the other: VMs provide higher isolation, security and portability, while processes are more efficient, flexible and easier to schedule and cooperate. Currently, this heavyweight architecture degrades both efficiency and security of cloud services. There are two trends for cloud virtualization: the first is to enhance processes to achieve VM-like security, and the second is to reduce VMs to achieve process-like flexibility. Based on these observations, our vision is that in the near future VMs and processes might be fused into one new abstraction for cloud virtualization that embraces the best of both, providing VM-level isolation and security while preserving process-level efficiency and flexibility. We describe a reference implementation, dubbed cKernel (customized kernel), for the new abstraction. Essentially, cKernel enhances the exokernel architecture by (i) adopting the LibOS paradigm to assemble isolated, smallest possible "execution environments", and (ii) following the the "core-shell" model to dynamically add traditional process features to the environments.

JointCloud computing is a new generation cloud computing model based on collaboration among Cloud Service Providers, making resources from multiple clouds deeply integrated., and supporting customize cloud service. To achieve the data confirmation Right when doing data trading to prevent resell is one of the most important challenges faced by such JointCloud environment. In this paper, we propose the JointCloud Computing Data Trading Architecture (JCDTA), an optimized data trading architecture for cross-stakeholder. Firstly, the announced mechanism is used to extract the data resources summary and the owner information into blockchain. Secondly, the untampering of the blockchain is used to record the data resources information, such as data resources statement information, data resources transaction records and data resources operation records. Finally, the supervision mechanism confirms the declaration of the data sources while the encryption technology ensures the privacy of the data resources. JCDTA guarantees the security and the reliability of data trading in the JointCloud environment and ensures the value invariance of data resources.

Complete coverage, which is the foundation of many robotic applications, aims to cover an area as quickly as possible. This study investigates the time-limited version of multi-robot complete coverage problem, that is, to find the least number of robots and allocate tasks properly to them such that they can finish a known mission within the time limit. This version of problem can be tackled straightforwardly based on optimizing the task-allocation to a fixed number of robots and enumerating the number. However, the number-fixed problem is NP-hard and the existing algorithm for the number-fixed problem allows intersecting tasks (possibly causing robots' interference) and endures high approximation factor. In this study, the time-limited complete coverage problem is tackled with a multi-objective approach, instead of enumerating robots' number and optimizing each number-fixed problem one by one. The multi-objective GA, Mofint, at first estimates the lower and upper bounds of the number of robots. It abstracts each task as a weighted node of a graph. Then, Mofint evolves individuals, each individual being a forest containing a certain number (within the bounds) of non-intersecting trees. Mofint can finally obtain higher precision than existing work with less time: the approximation factor for Mofint is 1.1 to 1.5 times the ideal allocation when robots' number is fixed, while for existing work is 1.5 to 2. Due to its higher precision, the least number of robots obtained in the experiments by Mofint is 0.6 times of existing work.

In the past decade, multi-robot simultaneous localization and mapping (SLAM) has been widely studied. However, the problem of collaborative SLAM with a large number of robots, such as dozens of robots, is far from being well solved. The challenges stem from not only the computation complexity in large-scale map merging but also the inefficiency to enable the parallel computing in this process, which is indispensable for us to make avail of the frontier of computing technology such as powerful cloud infrastructure. To effectively address these challenges, especially the latter one, we propose a scalable and real-time multi-robot visual SLAM framework based on the cloud robotic paradigm. The prominent feature of our framework is that it can distribute the SLAM process to multiple computing hosts in a cluster, which enables map building in parallel. To eliminate the bottleneck from data sharing between different sub-tasks, we also introduce diversified messaging pattern for various messaging scenarios, as well as the consistency policies for map data. The evaluations on the prototype of our framework, have shown that our method can do support as many as 256 robot entities simultaneously, without any compromising on the precision of poses estimation and map building.

Following trails in the wild is an essential capability of out-door autonomous mobile robots. Recently, deep learningbased approaches have made great advancements in this field. However, the existing research only focuses on the trail following with a single robot. In contrast, many robotic tasks in the reality, such as search and patrolling, are conducted by a group of robots. While these robots are grouped to move in the wild, they can cooperate to significantly promote the trail following accuracy, for example, by sharing images of different view angles or real-time decision fusion. This paper proposes such an approach named DL-Cooper that enables multi-robot visionbased trail following based on deep learning algorithms. It allows each robot to make a decision respectively with deep neural network and then fusion the decisions on the collective level with the support of back-end cloud computing infrastructure. It also takes Quality of Service (QoS) assurance, a very essential property of robotic software, into consideration. By limiting the condition to fusion decisions, the time latency can be minimally sacrificed. Experiments on the real-world dataset show that our approach has significantly improved the accuracy of the singlerobot system.

We present an unsupervised learning framework for the task of monocular depth and camera motion estimation from video sequences. In common with recent work, we use an unsupervised end-to-end learning method, requiring monocular video sequences for training. What makes the difference is, our approach not only uses image reconstruction as the supervisory signal but also exploits the pose estimation method which was used in traditional SLAM approach to enhance the supervisory signal and add training constraints. In pose estimation, a continuous frame window is set to construct the pose graph. Our method uses single-view depth and multi-view pose networks, with a loss based on reconstructing nearby images to the target using the predicted depth and pose. During training, the networks are thus coupled by the loss but can be applied independently at test time. Our evaluation of experiments on the KITTI dataset proves the effectiveness of our method: 1) monocular depth performs superior to the supervised methods that use ground-truth depth data for training and the existing unsupervised learning method. Our method performs comparably with the supervised methods that use ground-truth pose data for training. 2) pose estimation performs almost the same compared to established SLAM systems under comparable input settings.

The continuous contribution from peripheral participants is crucial for the success of open source projects. Thus, how to identify the potential Long-Term Contributors (LTC) early and retain them is of great importance. We propose a prediction model to measure the chance for an individual to become a LTC contributor through his capacity, willingness, and the opportunity to contribute at the time of joining. Using data of Rails hosted on GitHub, we find that the probability for a new joiner to become a LTC is associated with his willingness and environment. Specifically, future LTCs tend to be more active and show more community-oriented attitude than other joiners during their first month. This implies that the interaction between individual's attitude and project's climate are associated with the odds that an individual would become a valuable contributor or disengage from the project. We evaluated our prediction model by using the 10 cross-validation method. Results show that our model archives the mean AUC as 0.807, which is valuable for OSS projects to identify potential long-term contributors and adopt better strategies to retain them for continuous contribution.

In modern software engineering practices, various kinds of automated and intelligent methodologies have been proposed to improve the efficiency of collaborative development. However, most of those approaches are heavily dependent on supervised or semi-supervised learning technologies, which would be restricted by the lack of training data. Inspired by the theories and techniques of transfer learning, cross-project approaches have been proposed, but hard to achieve a consistent and desirable performances. In this paper, we conduct an extensive empirical study to capture the determinants that affect the performances of transferring reusable models across projects in the context of issue classification. Starting from a large-scale dataset, containing 799 OSS projects and more than 795,000 issues, we have extracted 28 attributes grouped into 4 different dimensions. The results show that the performance of cross-project issue classification based on model transferring is sensitive and unstable, which is influenced by multiple factors spreading among transferred model training, project construction, and technical and socail relations between source and target.

In GitHub, the pull-based development model enables community contributors to collaborate in a more efficient way. However, the distributed and parallel characteristics of this model pose a potential risk for developers to submit duplicate pull-requests (PRs), which increase the extra cost of project maintenance. To facilitate the further studies to better understand and solve the issues introduced by duplicate PRs, we construct a large dataset of historical duplicate PRs extracted from 26 popular open source projects in GitHub by using a semi-automatic approach. Furthermore, we present some preliminary applications to illustrate how further researches can be conducted based on this dataset.

Code summaries are human-readable text that describes the functionality of code blocks.Software developers use code summaries to understand the specification of API while code retrieve system relies on code summaries for effective code search.However, code summaries are often written by software developers.Writing good code summaries usually requires great effort.It could be helpful if developers use automatic code summarization system to generate code summaries.Recently, some works have applied deep learning methods to generate code summaries for code snippets.However, those deep learning methods treat code snippets as streams of text tokens while ignoring the inherent code structure information.In this paper, we propose a novel code summarization method named the CDE-Model (Code summarization by Deep learning and Empirical knowledge) that combines inherent code structure information with deep learning models.The CDE-Model proposes several empirical strategies to transform code snippets to refined code representation and feeds them into an encoder-decoder neural network for text generation.We conduct large-scale experiments on 1500 popular Java projects on GitHub 1 with 396,184 pairs of code snippets and summaries.Experimental results show that the quality of code summaries generated by our CDE-Model is better than other two methods.To the best of our knowledge, this paper is the first to combine code structure information with deep learning.

Continuous deployment (CD) is a software development practice aimed at automating delivery and deployment of a software product, following any changes to its code. If properly implemented, CD together with other automation in the development process can bring numerous benefits, including higher control and flexibility over release schedules, lower risks, fewer defects, and easier on-boarding of new developers. Here we focus on the (r)evolution in CD workflows caused by containerization, the virtualization technology that enables packaging an application together with all its dependencies and execution environment in a light-weight, self-contained unit, of which Docker has become the de-facto industry standard. There are many available choices for containerized CD workflows, some more appropriate than others for a given project. Owing to cross-listing of GitHub projects on Docker Hub, in this paper we report on a mixed-methods study to shed light on developers' experiences and expectations with containerized CD workflows. Starting from a survey, we explore the motivations, specific workflows, needs, and barriers with containerized CD. We find two prominent workflows, based on the automated builds feature on Docker Hub or continuous integration services, with different trade-offs. We then propose hypotheses and test them in a large-scale quantitative study.

No abstract available.

Collaborative development communities and knowledge sharing communities are highly correlated and mutually complementary. The knowledge sharing between these two types of open source communities can be very beneficial to both of them. However, it is a great challenge to automate this process. Current studies mainly focus on knowledge acquisition in one type of community, and few of them have tackle this problem efficiently. In this paper we take Android Issue Tracker and Stack Overflow as a case to study the mutual knowledge sharing between them. We propose an automatic approach by integrating semantic similarity with temporal locality between Android issues and Stack Overflow posts based on the internal citation-graph to reveal the potential associations between them. Our approach explores the internal citations in communities for closely related posts or issues clustering, exploits the rich semantics in fine-grained information of issues and posts for associations building, and leverages the temporal correlations between issues and posts in-depth for associations ranking. Extensive experiments show that the precision of our approach reaches 62.51 percent for top 10 recommendations when recommending Stack Overflow posts to Android issues, and 66.83 percent in reverse.

Many robotic tasks require heavy computation, which can easily exceed the robot's onboard computer capability. A promising solution to address this challenge is outsourcing thecomputation to the cloud. However, exploiting the potential ofcloud resources in robotic software is difficult, because it in-volves complex code modification and extensive (re)configurationprocedures. Moreover, quality of service (QoS) such as timeliness, which is critical to robot's behavior, have to be considered. Inthis paper, we propose a transparent and QoS-aware softwareframework called Cloudroid for cloud robotic applications. Thisframework supports direct deployment of existing robotic soft-ware packages to the cloud, transparently transforming theminto Internet-accessible cloud services. And with the automati-cally generated service stubs, robotic applications can outsourcetheir computation to the cloud without any code modification. Furthermore, the robot and the cloud can cooperate to maintainthe specific QoS property such as request response time, evenin a highly dynamic and resource-competitive environment. Weevaluated Cloudroid based on a group of typical robotic scenariosand a set of software packages widely adopted in real-worldrobot practices. Results show that robots capability can beenhanced significantly without code modification and specific QoSobjectives can be guaranteed. In certain tasks, the "cloud + robot" setup shows improved performance in orders of magnitudecompared with the robot native setup.

Currently, open source projects receive various kinds of issues daily, because of the extreme openness of Issue Tracking System (ITS) in GitHub. ITS is a labor-intensive and time-consuming task of issue categorization for project managers. However, a contributor is only required a short textual abstract to report an issue in GitHub. Thus, most traditional classification approaches based on detailed and structured data (e.g., priority, severity, software version and so on) are difficult to adopt. In this paper, issue classification approaches on a large-scale dataset, including 80 popular projects and over 252,000 issue reports collected from GitHub, were investigated. First, four traditional text-based classification methods and their performances were discussed. Semantic perplexity (i.e., an issues description confuses bug-related sentences with nonbug-related sentences) is a crucial factor that affects the classification performances based on quantitative and qualitative study. Finally, A two-stage classifier framework based on the novel metrics of semantic perplexity of issue reports was designed. Results show that our two-stage classification can significantly improve issue classification performances.

Cloud computing has completely changed the economics of IT industry. Recently, the new form of "shared global economy" requires cloud services to be collaboratively provisioned by different cloud providers in a Geo-distributed manner, which brings severe challenges in service performance and cost. To address this problem, in this paper we propose JointCloud, a cross-cloud cooperation architecture for integrated Internet service customization. JointCloud borrows the idea from airline alliances and aims at empowering the cooperation among multiple clouds to provide efficient cross-cloud services. JointCloud focuses not only on the vertical integration of cloud resources but also on the horizontal cooperation among different cloud vendors. This paper describes the concept and architecture of JointCloud, as well as the initial design of the key components, namely, communication, storage, and computation.

Uncertainty is a great challenge for environment perception of autonomous robots. For instance, while building semantic maps (i.e., maps with semantic labels such as object names), the robot may encounter unexpected objects of which it has no knowledge. It will lead to inevitable failures in traditional environment modeling software. The abundant knowledge being accumulated on the Internet has the potential to assist robots to handle such kind of uncertainly. However, existing researches have not touched this issue yet. This paper proposes a cloud-based semantic mapping engine named SemaCloud, which can not only augment robot's environment modeling capability by the rich cloud resources but also cope with uncertainty by utilizing the Internet knowledge on necessary. It adopts a state-of-art Deep Neural Network (DNN) for real-time and accurate recognition of pre-trained objects. If an object is beyond the knowledge of this DNN, a special mechanism named QoS-aware cloud phase transition is triggered to seek help from existing recognition services on the Internet. By a set of carefully-designed algorithms, it can maximize benefits and minimize the negative impacts on the Quality of Service (QoS) properties of robotic applications, which is essential to many robot scenarios. The experiments on both open datasets and real robots show that our work can handle uncertainly successfully in robotic semantic mapping without sacrificing critical real-time constraints.

The widespread use of pull-requests boosts the development and evolution for many open source software projects. However, due to the parallel and uncoordinated nature of development process in GitHub, duplicate pull-requests may be submitted by different contributors to solve the same problem. Duplicate pull-requests increase the maintenance cost of GitHub, result in the waste of time spent on the redundant effort of code review, and even frustrate developers' willing to offer continuous contribution. In this paper, we investigate using text information to automatically detect duplicate pull-requests in GitHub. For a new-arriving pull-request, we compare the textual similarity between it and other existing pull-requests, and then return a candidate list of the most similar ones. We evaluate our approach on three popular projects hosted in GitHub, namely Rails, Elasticsearch and Angular.JS. The evaluation shows that about 55.3% -- 71.0% of the duplicates can be found when we use the combination of title similarity and description similarity.

Code review is an important process to reduce code defects and improve software quality. However, in social coding communities using the pull-based model, everyone can submit code changes, which increases the required code review efforts. Therefore, there is a great need of knowing the process of code review and analyzing the pre-existing reviewer recommendation algorithms. In this paper, we do an empirical study about the PRs and their reviewers in Rails project. Moreover, we reproduce a popular and effective IR-based code reviewer recommendation algorithm and validate it on our dataset which contains 16,049 PRs. We find that the inactive reviewers are very important to code reviewing process, however, the pre-existing method's recommendation result strongly depends on the activeness of reviewers.

The pull-based model, widely used in distributed software development, allows any contributor to fork a public repository, package contributions as a pull-request, and then merge back to the original repository.Code review is one of the most significant stages in pull-based development.It ensures that only high-quality pull-requests are accepted, based on the in-depth discussion among reviewers.Thus, automatically identifying what reviewers are talking about in the discussions is benificial to better understand the code review process.In this paper, we conduct a case study on three popular opensource software projects hosted on GitHub and construct a finegrained taxonomy including 11 sub-categories for review comments.We then manually label over 5,600 review comments, and propose a Two-Stage Hybrid Classification (TSHC) algorithm to classify review comments automatically by combining rule-based and machine-learning techniques.Comparative experiments with a text-based method achieve a reasonable improvement on each project (9.2% in Rails, 5.3% in Elasticsearch, and 7.2% in Angular.jsrespectively) in terms of the weighted average Fmeasure.

The crowds' continuous participation and contribution are the key factors for the success of open source projects.However, among the massive competitors, it is difficult for a project to attract enough contributors by just passively waiting for enthusiasts to join in.Instead, it should actively seek gifted developers.Most of the current studies mainly focus on recommending experts inside a repository for some specific development tasks.In this paper, we propose a novel approach ConRec to recommend potential contributors across the entire open source community for given projects.It leverages the developers' historical activities in projects to analyze their technical interests and technical connections with others.Thereafter, it combines collaborative filtering algorithm with text matching algorithm to recommend proper developers.We conducted extensive experiments on 5,995 open source projects and 2,938,620 developers in GitHub.The results show that the proposed algorithm can recommend contributors to open source projects with the best performance of 63% in accuracy, and solve the cold start problem as well.

The Internet based cyber-physical world has profoundly changed the information environment for the development of artificial intelligence (AI), bringing a new wave of AI research and promoting it into the new era of AI 2.0. As one of the most prominent characteristics of research in AI 2.0 era, crowd intelligence has attracted much attention from both industry and research communities. Specifically, crowd intelligence provides a novel problem-solving paradigm through gathering the intelligence of crowds to address challenges. In particular, due to the rapid development of the sharing economy, crowd intelligence not only becomes a new approach to solving scientific challenges, but has also been integrated into all kinds of application scenarios in daily life, e.g., online-to-offline (O2O) application, real-time traffic monitoring, and logistics management. In this paper, we survey existing studies of crowd intelligence. First, we describe the concept of crowd intelligence, and explain its relationship to the existing related concepts, e.g., crowdsourcing and human computation. Then, we introduce four categories of representative crowd intelligence platforms. We summarize three core research problems and the state-of-the-art techniques of crowd intelligence. Finally, we discuss promising future research directions of crowd intelligence.

Social coding represents a new style of software production and creation paradigm, and demands for new technologies of software reuse. Many people searching for projects package, we can provide good reuse recommendation. In this paper, we focus on an interesting research topic of recommending software repositories to social developers, which is challenging because of two points: the first is how to get the interest contexts of developers; and the second is how to rank the repository candidates for recommendation properly. We propose RepoLike, a new approach for recommending repositories to developers by predicting their interests. RepoLike explores the developers' historical development activities and the social connections with other programmers, mines the technical features of repositories and the dependencies among them, and then combines both aspects to recommend most interesting and inspiring repositories to developers. The experiment results show that our approach can surprisingly recommend closely correlated repositories to developers, and the critical test results show that the recommendation performance is strongly impacted by the interest context model.

Software development is either creation activities that rely on human creativity and talents, or manufacturing activities that follow the engineering processes. This talk introduces a software development platform and ecosystem that combines the strengths of the above two aspects. Firstly, we propose the Trustworthy Software Model (TSM) containing tree novel models, i.e., lifecycle model, evidence model and evolution model. Based on the TSM, we designed and implemented TRUSTIE, which distinguishes itself by providing the software collaborative development service and the resource sharing service with the general support of trustworthiness-analysis tools. TRUSTIE enables crowd-oriented collaboration among internal development teams and the external crowds by combining the software creation and software manufacturing in one ecosystem.

Large and sparse datasets with a lot of missing values are common in the big data era, such as user behaviors over a large number of items. Classification in such datasets is an important topic for machine learning and data mining. Practically, naive Bayes is still a popular classification algorithm for large sparse datasets, as its time and space complexity scales linearly with the size of non-missing values. However, several important questions about the behavior of naive Bayes are yet to be answered. For example, how different mechanisms of data missing, data sparsity, and the number of attributes systematically affect the learning curves and convergence? In this paper, we address several common data missing mechanisms and propose novel data generation methods based on these mechanisms. We generate large and sparse data systematically, and study the entire AUC (Area Under ROC Curve) learning curve and convergence behavior of naive Bayes. We not only have several important experiment observations, but also provide detailed theoretic studies. Finally, we summarize our empirical and theoretic results as an intuitive decision flowchart and a useful guideline for classifying large sparse datasets in practice.

Large and sparse datasets with a lot of missing values are common in the big data era. Naive Bayes is a good classification algorithm for such datasets, as its time and space complexity scales well with the size of non-missing values. However, several important questions about the behavior of naive Bayes are yet to be answered. For example, how different mechanisms of missing, data sparseness and the number of attributes systematically affect the learning curves and convergence? Recent work in classifying large and sparse real-world datasets still could not address these questions mainly because the data missing mechanisms of these datasets are not taken into account. In this paper, we propose two novel data missing and expansion mechanisms to answer these questions. We use the data missing mechanisms to generate large and sparse data with various properties, and study the entire learning curve and convergence behavior of naive Bayes. We made several observations, which are verified through detailed theoretical study. Our results are useful for learning large sparse data in practice.

In IaaS cloud environments, peak memory demand caused by hotspot applications in Virtual Machine (VM) often results in performance degradation within and outside of this VM. Some solutions such as host swapping and ballooning for memory consolidation and overcommitment have been proposed. These solutions, however, have no help for addressing guest swapping issues inside VM. Even though host holds sufficient memory pages, guest OS is unable to utilize free pages in host directly due to the semantic gap between VMM and it. Our goal is to alleviate the performance degradation by decreasing disk I/O operations generated by guest swapping. Based on the insight analysis of behavioral features of guest swapping, we design HybridSwap, a distributed scalable framework which organize surplus memory in all hosts within data center into virtual pools for swapping. This framework builds up a synthetic swapping mechanism in a peer-to-peer way, which VM can adaptively choose suitable pools for swapping. We implement the prototype of HybridSwap and evaluate it with different benchmarks. The results demonstrate that our solution has the ability to promote the guest swapping efficiency indeed. Even in some cases, it shows 2-5 times of performance promotion compared with the baseline setup.

The massive amounts of open source software provide sufficient reusable resources for software development. Most of the OSS communities adopt a kind of categorization or tagging mechanism to organize the software. However, the categorization often too coarse, while the tags are flat and fail to capture the inter-relation among them. In this paper, we propose a novel approach to reveal the latent relations between tags and build a tag hierarchy to help locate resources. We firstly build a co-occurrence network, based on which we compare the connotations of tags and construct a preliminary hierarchy. Then we leverage the domain knowledge of category in SourceForge to optimize and improve the relations between tags. At the end, we demonstrate the effectiveness of the constructed tag hierarchy with quantitative evaluation, which suggest the validation of our approach.

As the rapid growth of open source software, how to choose software from many alternatives becomes a great challenge. Traditional ranking approaches mainly focus on the characteristics of the software themselves, such as qualities, security, reliable and so on. In this paper we investigate the market demands for software engineers, and propose a novel approach for ranking software by analyzing the market requirements for special software. At the same time we conclude the characteristics of software advertisements and analyze the reasons that why these situations emerge and tendency of software market requirements. As industries always need to balance several different factors for selecting software, the market demands can be a good indicator for ranking software and software evaluating. This paper provides quite a different perspective and some interesting inferences on software market requirements, and it can be a valuable supplement for traditional ranking methods, as well as software evaluating.

Recently, many researches propose that social media tools can promote the collaboration among developers, which are beneficial to the software development. Nevertheless, there is little empirical evidence to confirm that using @-mention has indeed a beneficial impact on the issues in GitHub. In this paper, we analyze the data from GitHub and give some insights on how @-mention is used in the issues (general-issues and pull-requests). Our statistical results indicate that, @-mention attracts more participants and tends to be used in the difficult issues. @-mention favors the solving process of issues by enlarging the visibility of issues and facilitating the developers' collaboration. In addition to this global study, our study also build a @-network based on the @-mention database we extract. Through the @-network, we can mine the relationships and characteristics of developers in GitHub's issues.

In bug tracking system, the high volume of incoming bug reports poses a serious challenge to project managers. Triaging these bug reports manually consumes time and resources which leads to delaying the resolution of important bugs. StackOverflow is the most popular crowdsourcing Q&A community with plenty of bug-related posts. In this paper, we explore the correlation between bug severity and the crowd attributes of linked posts. Two typical types of projects' bug repositories are studied here, e.g. Mozilla (user-centric project) and Eclipse (developer-centric project). Our results show that the bug severity is consistent with the crowd-based knowledge both in Mozilla and Eclipse, i.e. the linked posts of severe bugs have higher score etc. in StackOverflow than non-severe bugs. This interesting phenomenon inspires us that we can optimize the existing evaluation methods of bug severity by incorporating the crowd-based knowledge from a third-party in future.

The pull-based development model, enabled by git and popularised by collaborative coding platforms like Bit Bucket, Gitorius, and GitHub, is widely used in distributed software teams. While this model lowers the barrier to entry for potential contributors (since anyone can submit pull requests to any repository), it also increases the burden on integrators (i.e., Members of a project's core team, responsible for evaluating the proposed changes and integrating them into the main development line), who struggle to keep up with the volume of incoming pull requests. In this paper we report on a quantitative study that tries to resolve which factors affect pull request evaluation latency in GitHub. Using regression modeling on data extracted from a sample of GitHub projects using the Travis-CI continuous integration service, we find that latency is a complex issue, requiring many independent variables to explain adequately.

Software processes comprise many steps; coding is followed by building, integration testing, system testing, deployment, operations, among others. Software process integration and automation have been areas of key concern in software engineering, ever since the pioneering work of Osterweil; market pressures for Agility, and open, decentralized, software development have provided additional pressures for progress in this area. But do these innovations actually help projects? Given the numerous confounding factors that can influence project performance, it can be a challenge to discern the effects of process integration and automation. Software project ecosystems such as GitHub provide a new opportunity in this regard: one can readily find large numbers of projects in various stages of process integration and automation, and gather data on various influencing factors as well as productivity and quality outcomes. In this paper we use large, historical data on process metrics and outcomes in GitHub projects to discern the effects of one specific innovation in process automation: continuous integration. Our main finding is that continuous integration improves the productivity of project teams, who can integrate more outside contributions, without an observable diminishment in code quality.

Github facilitates the pull-request mechanism as an outstanding social coding paradigm by integrating with social media. The review process of pull-requests is a typical crowd sourcing job which needs to solicit opinions of the community. Recommending appropriate reviewers can reduce the time between the submission of a pull-request and the actual review of it. In this paper, we firstly extend the traditional Machine Learning (ML) based approach of bug triaging to reviewer recommendation. Furthermore, we analyze social relations between contributors and reviewers, and propose a novel approach to recommend highly relevant reviewers by mining comment networks (CN) of given projects. Finally, we demonstrate the effectiveness of these two approaches with quantitative evaluations. The results show that CN-based approach achieves a significant improvement over the ML-based approach, and on average it reaches a precision of 78% and 67% for top-1 and top-2 recommendation respectively, and a recall of 77% for top-10 recommendation.

Pull-request mechanism is an outstanding social development method in Git Hub. @-mention is a social media tool that deeply integrated with pull-request mechanism. Recently, many research results show that social media tools can promote the collaborative software development, but few work focuses on the impacts of @-mention. In this paper, we conduct an exploratory study of @-mention in pull-request based software development, including its current situation and benefits. We obtain some interesting findings which indicate that @-mention is beneficial to the processing of pull-request. Our work also proposes some possible research directions and problems of the @-mention. It helps the developers and researchers notice the significance of @-mention in the pull-request based software development.

Pull-Request (PR) is the primary method for code contributions from thousands of developers in GitHub. To maintain the quality of software projects, PR review is an essential part of distributed software development. Assigning new PRs to appropriate reviewers will make the review process more effective which can reduce the time between the submission of a PR and the actual review of it. However, reviewer assignment is now organized manually in GitHub. To reduce this cost, we propose a reviewer recommender to predict highly relevant reviewers of incoming PRs. Combining information retrieval with social network analyzing, our approach takes full advantage of the textual semantic of PRs and the social relations of developers. We implement an online system to show how the reviewer recommender helps project managers to find potential reviewers from crowds. Our approach can reach a precision of 74% for top-1 recommendation, and a recall of 71% for top-10 recommendation.

Issue resolution is a central task for software development. The efficiency of issue fixing largely relies on the issue report quality. Stack Overflow, which hosts rich and real-time posts about programming-specific problems, is a valuable external source of knowledge for issue resolution. In this paper, we present CrossLink, an analysis framework that automatically introduce related posts in Stack Overflow to issues in Android Issue Tracker. This helps developers to leverage the abundant and professional knowledge from massive programmers in Stack Overflow for issue resolution. CrossLink explores the semantic similarities as well as the temporal associations between the two types of repositories to recommend Stack Overflow posts to Android issues. The internal links in Stack Overflow are also employed to improve the linking accuracy. The experiments prove the effectiveness of CrossLink with precision of 62.51% for top-10 recommendations, which is significantly higher than the state-of-art method.

With the prosperity of the Open Source Software, various software communities are formed and they attract huge amounts of developers to participate in distributed software development. For such software development paradigm, how to evaluate the skills of the developers comprehensively and automatically is critical. However, most of the existing researches assess the developers based on the Implementation aspects, such as the artifacts they created or edited. They ignore the developers' contributions in Social collaboration aspects, such as answering questions, giving advices, making comments or creating social connections. In this paper, we propose a novel model which evaluate the individuals' skills from both Implementation and Social collaboration aspects. Our model defines four metrics from muti-dimensions, including collaboration index, technical skill, community influence and development contribution. We carry out experiments on a real-world online software community. The results show that our approach can make more comprehensive measurement than the previous work.

The large amounts of software repositories over the Internet are fundamentally changing the traditional paradigms of software maintenance. Efficient categorization of the massive projects for retrieving the relevant software in these repositories is of vital importance for Internet-based maintenance tasks such as solution searching, best practices learning and so on. Many previous works have been conducted on software categorization by mining source code or byte code, which are only verified on relatively small collections of projects with coarse-grained categories or clusters. However, Internet-based software maintenance requires finer-grained, more scalable and language-independent categorization approaches. In this paper, we propose a novel approach to hierarchically categorize software projects based on their online profiles across multiple repositories. We design a SVM-based categorization framework to classify the massive number of software hierarchically. To improve the categorization performance, we aggregate different types of profile attributes from multiple repositories and design a weighted combination strategy which assigns greater weights to more important attributes. Extensive experiments are carried out on more than 18,000 projects across three repositories. The results show that our approach achieves significant improvements by using weighted combination, and the overall precision, recall and F-Measure can reach 71.41%, 65.60% and 68.38% in appropriate settings. Compared to the previous work, our approach presents competitive results with 123 finer-grained and multi-layered categories. In contrast to those using source code or byte code, our approach is more effective for large-scale and language-independent software categorization.

The "Internetware" paradigm is fundamentally changing the traditional way of software development. More and more software projects are developed, maintained and shared on the Internet. However, a large quantity of heterogeneous software resources have not been organized in a reasonable and efficient way. Software feature is an ideal material to characterize software resources. The effectiveness of feature-related tasks will be greatly improved, if a multi-grained feature repository is available. In this paper, we propose a novel approach for organizing, analyzing and recommending software features. Firstly, we construct a Hierarchical rEpository of Software feAture (HESA). Then, we mine the hidden affinities among the features and recommend relevant and high-quality features to stakeholders based on HESA. Finally, we conduct a user study to evaluate our approach quantitatively. The results show that HESA can organize software features in a more reasonable way compared to the traditional and the state-of-the-art approaches. The result of feature recommendation is effective and interesting.

Performance diagnosis is labor intensive in production cloud computing systems. Such systems typically face many real-world challenges, which the existing diagnosis techniques for such distributed systems cannot effectively solve. An efficient, unsupervised diagnosis tool for locating fine-grained performance anomalies is still lacking in production cloud computing systems. This paper proposes CloudDiag to bridge this gap. Combining a statistical technique and a fast matrix recovery algorithm, CloudDiag can efficiently pinpoint fine-grained causes of the performance problems, which does not require any domain-specific knowledge to the target system. CloudDiag has been applied in a practical production cloud computing systems to diagnose performance problems. We demonstrate the effectiveness of CloudDiag in three real-world case studies.

In large-scale cloud computing systems, the growing scale and complexity of component interactions pose great challenges for operators to understand the characteristics of system performance. Performance profiling has long been proved to be an effective approach to performance analysis; however, existing approaches do not consider two new requirements that emerge in cloud computing systems. First, the efficiency of the profiling becomes of critical concern; second, visual analytics should be utilized to make profiling results more readable. To address the above two issues, in this paper, we present P-Tracer, an online performance profiling approach specifically tailored for large-scale cloud computing systems. P-Tracer constructs a specific search engine that adopts a proactive way to process performance logs and generates particular indices for fast queries; furthermore, PTracer provides users with a suite of web-based interfaces to query statistical information of all kinds of services, which helps them quickly and intuitively understand system behavior. The approach has been successfully applied in Alibaba Cloud Computing Inc. to conduct online performance profiling both in production clusters and test clusters. Experience with one real-world case demonstrates that P-Tracer can effectively and efficiently help users conduct performance profiling and localize the primary causes of performance anomalies.

In cloud computing systems, end-to-end request tracing approach is helpful for developers to understand the runtime behavior of user requests. Based on trace logs, we propose an approach to localize the abnormal methods that are the primary causes of performance problems. Our approach involves three steps: (1) cluster the user requests into different categories according to request call sequences and select major categories; (2) extract the principal methods that might be the causes of performance degradation; (3) pick out abnormal methods from those principal methods in each major category. We conduct four cases of performance degradations to validate our approach over a real-world enterprise-class cloud computing platform. The experimental results show that our approach can locate the prime causes of performance problems with low false-positive rate and false-negative rate.

Network virtualization is recognized as an effective way to overcome the ossification of the Internet. However, the virtual network mapping problem (VNMP) is a critical challenge, focusing on how to map the virtual networks to the substrate network with efficient utilization of infrastructure resources. The problem can be divided into two phases: node mapping phase and link mapping phase. In the node mapping phase, the existing algorithms usually map those virtual nodes with a complete greedy strategy, without considering the topology among these virtual nodes, resulting in too long substrate paths (with multiple hops). Addressing this problem, we propose a topology awareness mapping algorithm, which considers the topology among these virtual nodes. In the link mapping phase, the new algorithm adopts the k -shortest path algorithm. Simulation results show that the new algorithm greatly increases the long-term average revenue, the acceptance ratio, and the long-term revenue-to-cost ratio ( R/C ).

Increasing Internet business and computing footprint motivate server consolidation in data centers. Through virtualization technology, server consolidation can reduce physical hosts and provide scalable services. However, the ineffective memory usage among multiple virtual machines (VMs) becomes the bottleneck in server consolidation environment. Because of inaccurate memory usage estimate and the lack of memory resource managements, there is much service performance degradation in data centers, even though they have occupied a large amount of memory. In order to improve this scenario, we first introduce VM's memory division view and VM's free memory division view. Based on them, we propose a hierarchal memory service mechanism. We have designed and implemented the corresponding memory scheduling algorithm to enhance memory efficiency and achieve service level agreement. The benchmark test results show that our implementation can save 30% physical memory with 1% to 5% performance degradation. Based on Xen virtualization platform and balloon driver technology, our works actually bring dramatic benefits to commercial cloud computing center which is providing more than 2,000 VMs' services to cloud computing users.

Requirement Engineering (RE) has become an attractive field in both industry and academic. Many RE approaches have been presented in the past years to support eliciting, modeling, analyzing and specifying requirements of system to be built. However, requirement characteristics of kinds of systems like large scale software intensive systems pose several issues to requirements analysis and therefore challenge the extant RE approaches. This paper investigates a number of important metaphors in RE and proposes a novel RE approach that adopts capability as requirement metaphor. We discuss the requirements challenges coming from the changes of system-to-be and argue the necessity to introduce new abstraction and technology into RE to deal with the problems. The notions of capability and the reason to adopt capability as requirement metaphor are analyzed. The meta-model and framework of capability-based requirement engineering is proposed. A case is also studied in order to illustrate our approach. Capability as new abstraction in RE provides a new way to represent, analyze and tradeoff requirements.

Internet fundamentally changes the model of software development, the demands of software quality, and the process of software resource sharing. Internet- based environment for trustworthy software production is recognized as a key topic of software engineering in both academic and software industry. In this paper, the concepts and models of trustworthy software are introduced which dominate the design of Trustie environment. Trustie provides trustworthy software components sharing by an evolving software repository, and provides collaborative software development in a customizable development platform powered by a software production line framework. Finally the layered practices of research and application based on Trustie preliminarily demonstrate the effectiveness as well as the promising future of this environment.

Reputation systems provide a promising way to build trust relationships between users in distributed cooperation systems, such as file sharing, streaming, distributed computing and social network, through which a user can distinguish good services or users from malicious ones and cooperate with them. However, most reputation models mainly focus on evaluating the qualities of different services in one dimension, but care less about the preferences of different users. This paper proposes a personalized reputation model which provides each user a personalized trust view on others according to his preference. In our approach, we aggregate the users' preferences with collaborative filtering method and qualify it with user similarity which is integrated into the computing of reputation values. The experimental results suggest that our model can resist possible kinds of malicious behaviors efficiently.

Peer-to-Peer (P2P) systems receive growing acceptance, and the need of identifying the states of nodes appears increasingly in a variety of P2P based applications. In this paper, we propose Hermes, an algorithm to efficiently spread and maintain the states of all nodes in large scale systems. Hermes uses an improved push & pull style gossip process to spread the states of all nodes, and proposes a novel synopsis technique to maintain the states of all nodes locally at each node. Simulation results show that, in a N-node network with commonly accepted configurations, Hermes is 2 rounds faster than push & pull style gossip to spread the state of one node to the whole network; each node only needs to store a very small part of global state data to maintain the whole view of all nodes states in the network, which is updated dynamically with high accuracy.

In many cases, we would like to enhance the predefined adaptability of a running application, for example, to enable it to cope with a strange environment. To make such kind of runtime modifications is a challenging task. In existing engineering practices, the online policy upgrade approach just focuses on the modification of adaptation decision logic and lacks system-level means to assess the validity of an upgrade. This paper proposes a framework for adaptive software that supports the online reconfiguration of each concern in the “sensing-decision-execution” adaptation loop. To achieve this goal, our framework supports an architecture style which encapsulates adaptation concerns as software architecture elements. And then, it maintains a runtime architecture model to enable the dynamic reconfiguration of those elements as well as help to ensure the validity of a change. A third party can selectively add, remove or replace part of this model to enhance the running application's adaptability. We validated this framework by two cases extracted from real life.

Over its lifetime, adaptive software may have to deal with the environment not anticipated during the original development. In such cases, we should introduce new adaptive code, for example, to detect the strange contexts or update the out-of-date adaptation decision logic. This paper proposes an engineering approach facilitates this kind of post-delivery modifications based on software architecture techniques. Our approach introduces a component model separates different adaptation concerns (sensing, decision and execution) as different types of software architecture elements. The clear separation lays the foundation for the independent maintenance of each concern. And then, with the aid of a container supports the instantiation and run-time modification of the software architecture model, those concerns can be bound together without recompiling the whole software, even while it is running. Our approach enables the fine-grained, low-cost modifications of delivered adaptive software in the case that an unanticipated environment emerges.

With security researchers relying on the virtual machine (VM) in their analysis work, malware has a significant stake in detecting the presence of a VM to avoid executing its vicious behavior. But hiding a VM from malware by building a transparent virtual machine monitor (VMM) is fundamentally infeasible, as well as impractical from a performance and engineering standpoint. This paper proposes a new idea from another perspective: hiding the "real" machine from the VMM-aware malware. We propose a minimal VMM called MiniVMM which can migrate a booted OS, our protecting concern, to this VMM on demand. In our protection model, all the untrusted code, although having been verified by VMM-based malware detectors, should be executed in this migrated OS. Instead of building a transparent VMM, MiniVMM advisedly exposes the VMM fingerprints to prevent the computer against VMM-aware malicious programs by deceiving them into deactivating their destructive behavior by themselves. MiniVMM has two key features: dynamic OS migration and commodity VMM fingerprints emulation. Unlike existing VMM solutions, MiniVMM can make the protected OS transfer between VMM mode and native mode dynamically. MiniVMM can also emulate the fingerprints of prevalent VMMs to make the protected computer more like a "real" VM. MiniVMM might be deployed as a considerable complement of the existing VMM-based security approaches to make the native OSes immune to the VMM-aware malware.

Context awareness is one of the most fundamental issues in pervasive computing. In this paper, component based context model based on middleware architecture is proposed. Moreover, OWL-based context ontology for modelling context information to easily share and reuse context knowledge is presented. By giving fire alarm scenario for our prototype, the proposed component based context architecture can be deployed in different context-aware application and can provide a middleware support for context representation and knowledge sharing.