Heterogeneous graph learning aims to generate meaningful node representations for graph-structured data with diverse node types and complex relations, facilitating downstream tasks such as node classification and clustering. However, existing methods often emphasize either coarse-grained relational structures or fine-grained node attributes, paying limited attention to the other, which constrains their ability to fully capture the intricate interplay between nodes and relations. To address this limitation, we propose a novel Granular Interaction Heterogeneous Graph Auto-Encoder (GIHGAE), which effectively balances granular fusion and interactions in heterogeneous graph learning. Specifically, GIHGAE employs a relation-level encoder as the primary structure extractor to capture coarse-grained relational dependencies across the graph. Complementarily, we design a node-level encoder that integrates fine-grained contextual details from diverse node attributes, refining representations. These multi-granular features are fused into holistic node embeddings. Additionally, to ensure seamless integration of fine-grained and coarse-grained information, we introduce a global-level decoder to model interactions between nodes and relations explicitly. Finally, to further enhance GIHGAE, we incorporate a dual-loss mechanism, combining reconstruction loss for feature preservation and prediction loss to enhance downstream task performance. Extensive experimental evaluations in heterogeneous graph learning tasks highlight the strong performance of GIHGAE, which consistently outperforms current state-of-the-art methods in classification accuracy, clustering quality, and link prediction performance.

Intrusion detection is a crucial method for addressing the security risks of the Industrial Internet of Things (IIoT). However, acquiring substantial and high-quality training data can be challenging for centralized schemes. While federated learning has shown great application prospects as a secure distributed solution, it also encounters the problems of heterogeneous and imbalanced data in real-world production environments. In this article, we propose a personalized federated learning scheme based on Shannon entropy metric (PFLSE), aimed at providing a high-accuracy customized detection model for local organizations. This scheme introduces Shannon entropy into the aggregation mechanism, allowing the edge agent model, which contains richer global information, to carry greater weight in the aggregation process. In the local training process, a two-stage training strategy based on the concept of personalized layer is firstly applied to strengthen the global features and local personalized representations. Secondly, considering the differential balance degree between various edge agent data, a Shannon entropy based dynamic loss function (SDL) is proposed, which combines focal loss and cross-entropy loss, to improve training stability and alleviate the difficulty of training on imbalanced data. Finally, a comprehensive experiment simulating a real-world environment shows that PFLSE exhibits reliable intrusion detection performance across metrics such as accuracy, precision, andF1-Score. Furthermore, it outperforms other methods in the scenarios involving non-independent and identically distributed (non-IID) data.

Nighttime rainy conditions severely degrade visual quality in applications such as autonomous driving and aerial surveillance, where images suffer from compounded low-light and rain degradations. Diffusion models offer strong generative priors but face limitations in image restoration, including poor controllability, structural distortion, and domain gaps with degraded images. We present MR-SDformer, a novel framework that integrates Retinex-based decomposition with diffusion priors for joint nighttime deraining and low-light enhancement. The key innovation is the Maximum-Value Retinex decomposition, which isolates high-intensity rain streaks into the illumination map and produces a rain-free reflectance map that faithfully preserves intrinsic scene content. This decomposition not only bridges the gap between rainy inputs and rain-free priors but also provides complementary guidance to the generative process. Building on this, we design an asymmetric Hybrid Conditional Transformer that leverages the decomposed illumination and reflectance maps to condition the frozen diffusion model more effectively, enabling precise multi-modal feature fusion and high-fidelity reconstruction. Extensive experiments on both synthetic and real-world datasets confirm that MR-SDformer achieves state-of-the-art performance, delivering clearer structure, enhanced illumination, and more realistic visual quality under nighttime rainy conditions. • Maximum-Value Retinex decouples rain and low-light for robust restoration. • Bridges domain gap to unlock frozen diffusion priors for multi-degradations. • Retinex-based attention injects physical priors into the diffusion process. • Achieves SOTA performance on both synthetic and real-world datasets.

API tutorials and Stack Overflow (SO) are crucial API learning resources. API tutorials help developers understand API usage in general contexts, while SO explains API usage in specific programming tasks. Using both API tutorials and SO provides more API knowledge. We treat a tutorial fragment or a SO Question and Answering (Q&A for short) pair as a knowledge item (KI for short). Discovering relevant KIs of the APIs helps developers understand and learn how to use APIs. However, existing relevant KIs identification approaches mainly focus on either API tutorials or SO. Furthermore, these approaches do not take into account code in KIs. In this paper, we propose PTIRK, a novel approach using prompt tuning to identify relevant KIs of API from both API tutorials and SO. Firstly, we extract textual descriptions and code snippets from KIs and match them with the API class name. Then, prompt tuning is used to tune BERT for text in KIs and CodeBERT for code in KIs. After that, we combine tuned models by using a gating fusion strategy. Finally, the relevant KIs of API can be identified by the trained model. We evaluate PTIRK on Java and Android datasets with 10,072 samples. Experimental results show that PTIRK outperforms state-of-the-art approaches on both datasets, and the user study further confirms its practical effectiveness.

Domain adaptation (DA) is a frequently used technique in intelligent fault diagnosis. However, existing DA methods presume that the source and target domains have the same label space. Due to the complexity of industrial operation conditions, new fault types will inevitably occur. Thus, the above assumption is only sometimes satisfied. To overcome this issue, we propose a novel Frequency-Aware Model with Neighborhood Invariance (FAN) for cross-domain open-set fault diagnosis. Firstly, we comprehensively consider the domain shift phenomenon in time and frequency features and construct an encoder based on the Fourier Neural Operator (FNO) to extract potential invariant information efficiently. Secondly, we expect known class samples to be mapped to an invariant neighborhood to separate unknown classes. Based on this, we adopt neighborhood invariance learning to reduce the intra-domain variations in the target domain and form robust discriminative boundaries. Extensive experiments on public and real-world datasets demonstrate that FAN outperforms the comparison methods and has flexibility.

With the rapid development of the blockchain industry and the widespread adoption of IoT devices, which are often deployed on different blockchains, the need for cross-chain value and data exchange has become increasingly important. However, existing cross-chain transactions face challenges such as low efficiency, high costs, and insufficient security. To address these issues, this paper proposes a cross-chain transaction scheme based on aggregated zero-knowledge proofs. This scheme optimizes the allocation of computing resources in a distributed environment and employs a multi-branch balanced Merkle tree to construct aggregated zero-knowledge proofs, significantly reducing the verification costs for batch cross-chain transactions.To further enhance data privacy and integrity, this paper introduces the Secure Aggregated Block Verification (SABV) algorithm and improves system consistency and reliability through the Local Merkle Tree Rebalance (LMTR) algorithm. In addition, this paper analyzes the basic security of the proposed scheme when implemented in adversarial environments and provides countermeasures for common threats in distributed systems. Finally, simulations and actual deployment on the Ethereum test network were conducted. The results indicate that our method reduces CPU usage, memory consumption, and time expenditure by 50.10%, 99.03%, and 99.47%, respectively, during the generation of zero-knowledge proofs for batch cross-chain transactions. At the same time, building upon the performance improvements of the existing zero-knowledge proofs, our approach also demonstrates significant enhancements in contract deployment and cross-chain transaction efficiency.

While Federated Learning is intended to safeguard data privacy, it is confronted with the problem of gradient leakage, which empowers attackers to execute gradient inversion attacks and retrieve the original data through the shared gradient information. Existing gradient inversion attack methods can achieve good results when handling small batches of low-resolution images. However, when dealing with large batches of high-resolution images, problems such as gradient ambiguity and model instability will occur, resulting in a significant decrease in the recovery performance. We propose a novel Image-correction-penalty based Over-parameterized Regression Network (IORN). IORN breaks through the limitations of existing methods with its unique design. The Adaptive Over-parameterized Network in IORN can dynamically adjust its structure, thereby enhancing the network’s ability to capture complex data distributions. This enables it to better handle the complexity of large batches of high-resolution images and improves the model’s reconstruction ability for such images. Meanwhile, the designed image correction penalty term restricts the difference between the generated images and the average image. This not only improves the stability of the optimization process but also reduces the convergence deviation. Experimental results demonstrate that IORN significantly improves the resolution and fidelity of reconstructed images during gradient inversion attacks on the MNIST, CIFAR-100, and LFW datasets, especially showing outstanding performance when dealing with large batches of complex images.

The surge in the development of artificial intelligence has led to increases in the complexity of computational tasks and the resource demands within cloud computing scenarios. Therefore, intelligent scheduling methods have formed a crucial research area. Solving complex scheduling problems requires many problem feature and long-sequence decision-making observations as possible. To address the workflow scheduling problem under the limited capabilities of models, workflow reduction and cross-view workflow scheduling problems are first proposed in this paper, with the optimization objectives and constraints of each problem described. Second, a cross-view intelligent scheduling method implemented via cloud computing workflow reduction (CSCR), including a workflow reduction sorting algorithm (Task-priority ranker), an intelligent reduction algorithm (Workflow view-transformer), and a cross-view intelligent scheduling algorithm (Joint-scheduler), is proposed. We also propose an intelligent scheduling architecture under the workflow reduction paradigm. By reducing the workflow, we provide multiple views that support the decision-making processes of deep reinforcement learning-based scheduling models and coordinate workflow views before and after the reduction step to achieve cross-view joint scheduling. Experimental results show that CSCR achieves minimum advantages of 42.1%, 43.2%, and 33.3% in terms of three workflow reduction indicators over four other algorithms, significantly optimizing the effect of the employed scheduling model.

The in-depth development of artificial intelligence models has supported the high-quality allocation of cloud computing resources. The optimization of workflow scheduling issues in cloud computing has become increasingly critical due to the complexity of computing tasks, constraints on computing resources, and the growing demand for high-quality service. To address the increasingly complex workflow scheduling problems in cloud computing, this paper presents an arbitrary-freedom adaptive scheduling method for cloud computing with multiple workflows based on deep reinforcement learning (termed AFAS), with the workflow makespan and response time as the optimization objectives. First, we define the concept of degrees of freedom in the scheduling context to establish the feature space and foundational decision patterns relevant to multiworkflow scheduling. Second, an adaptive real-time scheduling strategy generation (ARS) algorithm is proposed for multiworkflow scheduling tasks. Third, a composite reward mechanism with an advanced-time-window real-time-reward (ATR) algorithm is designed for intelligent model optimization. Finally, the generation algorithm and intelligent model are fused to perform arbitrary-freedom multiworkflow adaptive scheduling. The experiments show that ATR can significantly increase the frequency of reward generation, AFAS can achieve at least 6.6% better performance than existing methods can achieve, and the incorporation of intelligent models improves the performance of ARS by 2.7%.

Light-detection-and-ranging-based multiobject detection and tracking play fundamental roles in autonomous driving systems. Most existing detection and tracking methods inevitably require complex pairing permutations for object association across frames, making the framework slow. Moreover, the occlusion and viewpoint changes lead to missed and false detection. To solve the abovementioned issues, this article proposes a simple joint 3-D detection and tracking approach with three-step offset learning ($\text{Offset}^{3}\text{Net}$). Specifically,$\text{Offset}^{3}\text{Net}$incorporates three task-specific output subnetworks to learn three offsets: 1) center offset, 2) motion offset, and 3) association offset. The learning of abovementioned offsets eliminates the complex bipartite matching processing. Specifically, the center offset guides the model to generate precise detections, whereas the motion offset transforms the track from the previous frame to the current frame, and the association offset minimizes the distance between detection and motion-updated track of the same object. Then, a simple read-off operation is conducted for data association on a hybrid-time centerness map, which represents the detections and offset-updated tracks. In addition, we design a detection-feature-enhanced module that captures the temporal coherence of the object motion and appearance information, avoiding the missed and false detection. Experiments on nuScenes have demonstrated the effectiveness of our$\text{Offset}^{3}\text{Net}$in terms of accuracy and speed compared with most 3-D detection and tracking methods.

Deep learning has shown broad research prospects in addressing insider threats, a serious problem currently facing industrial information systems. Although deep learning is able to capture effective feature representations from complex multidimensional data, there are still issues such as strong stealth of insider threat behavior and the imbalance data that need to be solved. Therefore, we propose an adversarial Autoencoder based Unsupervised insider Threat detection scHeme (AUTH). Compared to other methods, AUTH fully considers the role of time feature and event feature in threat detection. In addition, in order to improve the performance of autoencoder models to detect covert threat behaviors, AUTH drives a temporal convolutional network and long short-term memory network-based Adversarial Autoencoder (TL-AAE). Generative Adversarial Theory is introduced to solve the problem of uncertainty in the latent feature of the encoder. Finally, with the sufficient experiments on public datasets, we demonstrate that the usefulness of adding time features and the proposed TL-AAE model to improve threat detection performance. Compared with the baseline, AUTH obtains the area under curve value of 0.932, which is 4.95% higher than the highest result obtained by the baseline. In addition, AUTH obtains the EER value of 0.146, which is 12.57% lower than the lowest result of the baseline.

With the vigorous development of the blockchain industry, cross-chain transactions can effectively solve the problem of “islands of value” caused by the inability to interact between different chains. However, security risks in reputation management caused by cross-chain transactions implemented through notary solutions have always existed. Consequently, this paper proposes a blockchain cross-chain transaction method based on decentralized dynamic reputation value assessment. The notary election phase addresses the issue of the continually changing behaviour of notaries in actual transactions by designing a dynamic evaluation window mechanism based on an RNN. Moreover, a reputation-rating decay mechanism is introduced to avoid the problem of reputation value recovery caused by malicious notaries being inactive for a long time. Relative to alternative reputation assessment models, the proposed method offers a thorough evaluation of user behavior and effectively identifies malicious activities in real-time. Finally, the method was tested by deploying it on the Ethereum blockchain. Our approach offers more dynamic settings for window parameters, adapting to changes in notary behavior and reducing the number of detections within the same timeframe by approximately 59.14%. The weight factor settings are also optimized, allowing for adjustments based on specific situations to achieve accurate reputation values. Overall, this method not only enhances the security of cross-chain transactions but also reduces operational costs by 53.3% compared to traditional technologies.

Convolutional neural network (CNN) accelerator has been gradually deployed on edge Field-Programmable Gate Arrays (FPGAs) for satellite remote sensing. However, the considerable complexity of software/hardware co-design inhibits the development of data processing applications on edge FPGAs. Moreover, the performance of edge FPGAs is restricted due to limited bandwidth and excessive software/hardware communication overhead in satellite remote sensing. To reduce co-design complexity, we propose a software/hardware data mapping framework for the deployment efficiency of the remote sensing data accelerator, called Fast-Pipe. Especially the software design of Fast-Pipe takes into account the interaction requirements of accelerator deployment and the poll and copy latency of data transfers. The data transfer driver is implemented in the user space to reduce latency with a new scheduling method and an interrupt policy for sending and receiving data. Meanwhile, an associated hardware structure is designed with Direct Memory Access (DMA) to implement data scheduling and mapping. With Fast-Pipe, the data in the software buffer is mapped into multiplexed data streams. Experimental results show that Fast-Pipe speed up 46.38× in small data transfer, and average 2× higher in large data transfer than previous work with stable data transfer speed.

Aiming at the problems of computing power, reliability and cost when intelligent vehicles deal with computationally intensive and delay-sensitive emerging applications in multiple business scenarios in the Internet of Vehicles, an intelligent computing offloading model is proposed. This can minimize the total system cost under the constraint of time delay and energy consumption. Considering the cost of blockchain and the cost of intelligent vehicles, the DDPG algorithm is used to solve the proposed model. Simulation results show that the method proposed in this paper can effectively reduce the total cost of computing offload and further improve the success rate of computational offloading under the premise of computational offloading safety.

This paper presents a novel self-bias CMOS rectifier for ultra-high frequency (UHF) RF energy harvester, which achieves a high power conversion efficiency (PCE) at an ultra-low input power. Compared to other state-of-the-art topologies, the proposed architecture combines static and dynamic compensated techniques to enhance the conduction capacity when the input power is lower than -27dBm. A six- stage rectifier is designed using this proposed topology in a 130nm CMOS process technology. Operating at 915MHz and driving a 1MΩ load resistor, the post-layout simulation PCE of this work is 27.6% at -30dBm input power. A sensitivity of - 30dBm is stimulated with 0.8V output voltage across a capacitive load.

Abstract Ultra‐dense networking is considered an effective solution to achieve high capacity in 5G networks. However, the densely distributed base stations (BSs) in ultra‐dense networks (UDNs) make the inter‐cell interference much more serious than that in traditional cellular networks. Therefore, it is important to mitigate inter‐cell interference in the UDNs to improve network performance. To tackle this problem, we propose a novel virtual resource mapping algorithm that includes a resource reservation (RR) algorithm and a real‐time resource embedding (RE) algorithm. Specifically, according to the number of services predicted by a dynamic service model, the RR algorithm is proposed to determine the sets of multiplexing BSs in the next time cycle and reserve channel resource required by each BS. Then, to further reduce inter‐cell interference, the real‐time RE algorithm is proposed to allocate the channel resource in real time. Finally, simulation results show that the proposed algorithm has better performance in terms of signal‐to‐interference‐plus‐noise ratio and acceptance ratio, compared to the existing algorithms, such as the frequency reuse channel allocation algorithm and inter‐cell interference coordination algorithm.

The heterogeneous defect prediction (HDP) technique can predict defects in a target company using heterogeneous metric data from external company, which has received substantial research attention. However, existing HDP methods assume that source data is labeled but labeling data is expensive. Semi-supervised defect prediction technique can perform defect prediction with few labeled data. In this paper, we investigate a new problem — semi-supervised HDP (SHDP). To solve this problem, we propose a new approach named cost-sensitive kernel semi-supervised correlation analysis (CKSCA) as a solution of SHDP problem. It introduces unified metric representation and canonical correlation analysis to make the data distributions of different company projects more similar. CKSCA also designs a cost-sensitive kernel semi-supervised discriminant analysis mechanism to utilize the limited labeled data and sufficient real-life unlabeled data from different companies. Besides we collect lots of open-source projects from GitHub website to construct a new large-scale unlabeled dataset called GITHUB dataset. It contains 26,407 modules and is greater than each public project dataset. It has been public online and can be extended continuously. Experiments on the GITHUB dataset and other public datasets indicate that unlabeled GITHUB data can help prediction model improve prediction performance, and CKSCA is effective and efficient for solving SHDP problem.

The parameter server architecture has shown promising performance advantages when handling deep learning (DL) applications. One crucial issue in this regard is the presence of stragglers, which significantly retards DL training progress. Previous solutions for solving stragglers may not fully exploit the computation resource of the cluster as evidenced by our experiments, especially in the heterogeneous environment. This motivates us to design a heterogeneity-aware parameter server paradigm that addresses stragglers and accelerates DL training from the perspective of computation parallelism. We introduce a novel methodology named straggler projection to give a comprehensive inspection of stragglers and reveal practical guidelines to solve this problem in two aspects: (1) controlling each worker's training speed via elastic training parallelism control and (2) transferring blocked tasks from stragglers to pioneers to fully utilize the computation resource. Following these guidelines, we propose the abstraction of parallelism as an infrastructure and design the Elastic-Parallelism Synchronous Parallel (EPSP) algorithm to handle distributed training and parameter synchronization, supporting both enforcedand slack-synchronization schemes. The whole idea has been implemented into a prototype called Falcon which effectively accelerates the DL training speed with the presence of stragglers. Evaluation under various benchmarks with baseline comparison demonstrates the superiority of our system. Specifically, Falcon reduces the training convergence time, by up to 61.83, 55.19, 38.92, and 23.68 percent shorter than FlexRR, Sync-opt, ConSGD, and DynSGD, respectively.

The multi-interface networks are efficient infrastructures to deploy distributed Deep Learning (DL) tasks as the model gradients generated by each worker can be exchanged to others via different links in parallel. Although this decentralized parameter synchronization mechanism can reduce the time of gradient exchange, building a high-performance distributed DL architecture still requires the balance of communication efficiency and computational utilization, i.e., addressing the issues of traffic burst, data consistency, and programming convenience. To achieve this goal, we intend to asynchronously exchange gradient pieces without the central control in multi-interface networks. We propose the Piece-level Gradient Exchange and Multi-interface Collective Communication to handle parameter synchronization and traffic transmission, respectively. Specifically, we design the gradient sketch approach based on 8-bit uniform quantization to compress gradient tensors and introduce the colayerabstraction to better handle gradient partition, exchange and pipelining. Also, we provide general programming interfaces to capture the synchronization semantics and build the Gradient Exchange Index (GEI) data structures to make our approach online applicable. We implement our algorithms into a prototype system called Canary by using PyTorch-1.4.0. Experiments conducted in Alibaba Cloud demonstrate that Canary reduces 56.28 percent traffic on average and completes the training by up to 1.61x, 2.28x, and 2.84x faster than BML, Ako on PyTorch, and PS on TensorFlow, respectively.

As to address the impact of heterogeneity in distributed Deep Learning (DL) systems, most previous approaches focus on prioritizing the contribution of fast workers and reducing the involvement of slow workers, incurring the limitations of workload imbalance and computation inefficiency. We reveal that grouping workers into communities, an abstraction proposed by us, and handling parameter synchronization in community level can conquer these limitations and accelerate the training convergence progress. The inspiration of community comes from our exploration of prior knowledge about the similarity between workers, which is often neglected by previous work. These observations motivate us to propose a new synchronization mechanism named Community-aware Synchronous Parallel (CSP), which uses the Asynchronous Advantage Actor-Critic (A3C), a Reinforcement Learning (RL) based algorithm, to intelligently determine community configuration and fully improve the synchronization performance. The whole idea has been implemented in a system called Petrel that achieves a good balance between convergence efficiency and communication overhead. The evaluation under different benchmarks demonstrates our approach can effectively accelerate the training convergence speed and reduce synchro-nization traffic.

Cross‐project defect prediction (CPDP) technology refers to the constructing prediction model to predict the instance label of the target project by utilising labelled data from an external project. The challenge of CPDP methods is the distribution difference between the data from different projects. Transfer learning can transfer the knowledge from the source domain to the target domain with the aim to minimise the domain difference between different domains. However, most existing methods reduce the distribution discrepancy in the original feature space, where the features are high‐dimensional and non‐linear, which makes it hard to reduce the distribution distance between different projects. Moreover, previous works mainly consider marginal distribution or conditional distribution difference. In this study, the authors proposed a manifold embedded distribution adaptation (MDA) approach to narrow the distribution gap in manifold feature subspace. MDA maps source and target project data to manifold subspace and then joint distribution adaptation of conditional and marginal distributions is performed on manifold subspace. To evaluate the effectiveness of MDA, the authors perform extensive experiments on 20 public projects with three indicators. The experiment results show that MDA improves the average performance, but the improvement is not statistically significant in comparison to HYDRA (one of the baselines).

In edge-enabled Internet of Things (IoT), computation offloading service is expected to offer users with better Quality of Experience (QoE) than traditional IoT. Unfortunately, the growing multiple tasks from users are occuring with the emergence of the IoT environment. Meanwhile, the current computation offloading with QoE is solved by deep reinforcement learning (DRL) with the issue of instability and slow convergence. Therefore, improving the QoE in edge-enabled IoT is still the ultimate challenge. In this article, to enhance the QoE, we propose a new QoE model to study the computation offloading. Specifically, the emerged QoE model can capture three influential elements: 1) service latency determined by local computing latency and transmission latency; 2) energy consumption according to local calculation and transmission consumption; and 3) task success rate based on the coding error probability. Moreover, we improve the deep deterministic policy gradients (DDPG) algorithm and propose a algorithm named the double-dueling-deterministic policy gradients (D3PG) based on the proposed model. Specifically, the actor network highly relies on the critic network, which makes the performance of the DDPG sensitive to the critic and thus leads to poor stability and slow convergence in the computation offloading process. To solve this, we redesign the critic network by using Double Q -learning and Dueling networks. Extensive experiments verify the better stability and faster convergence of our proposed algorithm than existing methods. In addition, experiments also indicate that our proposed algorithm can improve the QoE performance.

With the rapid development of society and the economy and the increasing seriousness of environmental problems, renewable energy and high-quality energy services in low-carbon communities have become popular research topics. However, a large number of volatile distributed generation power systems in the community are connected to the grid. It is difficult to stabilize and efficiently interact with fragmented and isolated energy management systems, and it is difficult to meet energy management needs in terms of low-carbon emissions, stability, and intelligence. Therefore, by considering operation costs, pollution control costs, energy stability, and plug-in hybrid electric vehicles, this article proposes a regional energy supply model called community energy Internet and builds a low-carbon community energy adaptive management model for smart services. Then, to address energy supply instability, an adaptive feedback control mechanism developed based on model predictive control is introduced to adapt to the changing environment. Finally, a long short-term memory-recurrent neural network-based Tabu search is introduced to prevent the multiobjective particle swarm optimization algorithm from easily falling into a local optimum. The simulation results show that the proposed model can effectively realize the optimal allocation of energy, which solves the problem of fragmented energy islands caused by distributed power access. This method has quality of service benefits for users, such as cost, time, and stability, and realizes wide interconnections, high intelligence, and low-carbon efficiency of community energy management.

Parameter server paradigm has shown great performance superiority for handling deep learning (DL) applications. One crucial issue in this regard is the presence of stragglers, which significantly retards DL training progress. Previous solutions for solving straggler may not fully exploit the computation capacity of a cluster as evidenced by our experiments. This motivates us to make an attempt at building a new parameter server architecture that mitigates and addresses stragglers in heterogeneous DL from the perspective of computation parallelism. We introduce a novel methodology named straggler projection to give a comprehensive inspection of stragglers and reveal practical guidelines for resolving this problem: (1) reducing straggler emergence frequency via elastic parallelism control and (2) transferring blocked tasks to pioneer workers for fully exploiting cluster computation capacity. Following the guidelines, we propose the abstraction of parallelism as an infrastructure and elaborate the Elastic-Parallelism Synchronous Parallel (EPSP) that supports both enforced-and slack-synchronization schemes. The whole idea has been implemented in a prototype called Falcon which efficiently accelerates the DL training progress with the presence of stragglers. Evaluation under various benchmarks with baseline comparison evidences the superiority of our system. Specifically, Falcon yields shorter convergence time, by up to 61.83%, 55.19%, 38.92% and 23.68% reduction over FlexRR, Sync-opt, ConSGD and DynSGD, respectively.

In the era of information, the services of fog computing environment with content delivery are expected to offer users the better satisfaction of Quality-of- Experience (QoE) than that in a conventional environment. Nevertheless, the dataflow and new demands from users increase along with the promising of content-centric computing system in fog computing environment. Therefore, the satisfaction of QoE will become the major challenge. In this article, to enhance the satisfaction of QoE, we propose QoE models to evaluate the quality of service in fog computing environment concerning both system and users. The value of QoE does not only refer to the system cost, but also the Mean Opinion Score (MOS) of users. Therefore, our models could capture influential factors from system cost based on system states and services for users. Specially, we mainly focus on issues of cache allocation and transmission rate. Under this fog computing environment, aiming to the capacity of cache allocation among fog nodes and handle transmission rates under a constrained total system cost and MOS, we devote our efforts to the following two aspects. First, we formulate the QoE as a joint resource allocation problem under different transmission rates to acquire best QoE. Then, we propose a dynamic algorithm based on shortest path tree (SPT), which is suitable for fog computing environment with content delivery frequently. Simulation results reveal that the benefit for using the dynamic allocation (DA) method to allocate resource can achieve high QoE performance.

Fuel cells are promising power sources for green data centers thanks to its high energy-efficiency, low greenhouse gas emissions and high reliability. However, fuel cells have a unique feature called limited load following, i.e., they are slow in adjusting power supply due to mechanical limitation of fuel delivery. When power demand of data centers suddenly grows, fuel cells would fail to provide sufficient power supply. On the other hand, fuel cells are slow to reduce its power supply when demand decreases, leading to energy waste. In this paper, we study to mitigate the impact of limited load following by associating a set of heterogeneous batteries with fuel cells. These batteries with different characteristics (e.g., capacity, charging and discharging rate) can power data centers when the energy supply of fuel cells is insufficient. They are charged by excessive power supply when demand decreases. Given future power demand, we formulate the energy management problem as a mixed-integer nonlinear programming. An online algorithm is designed to solve the problem without future knowledge. We conduct extensive simulations using real-world traces and results show that our proposed algorithm significantly outperforms existing solutions.

Rapid economic development has been observed worldwide, which has caused environmental problems to worsen. Thus, the Energy Internet (EI), which accesses renewable energy and provides high-quality power services, has recently become a hot issue. As a subnet of the EI, an energy local area network (ELAN) consists of renewable power generation equipment, controllable distributed power generation equipment, storage systems, electric vehicles, and a large number of loads. Energy management is required for economic, environmental, and safety considerations. This paper proposes an energy management optimization model that addresses ELAN operations and includes pollution treatment fees; this model provides intelligent control of the charging and discharging of plug-in hybrid electric vehicles (PHEVs). This model achieves a nonlinear energy management optimization for an ELAN. To promote optimal performance, an improved comprehensive learning particle swarm optimization (CLPSO) algorithm is presented; it combines Tabu Search (TS) and CLPSO to avoid local optima. To verify the performance of our model, two experimental scenarios are built. The simulation results show that our energy management optimization model fulfills the optimal allocation of energy and that the PHEV intelligent charging/discharging strategy promotes economic benefits for the network.

When there exist not enough historical defect data for building an accurate prediction model, semisupervised defect prediction (SSDP) and cross-project defect prediction (CPDP) are two feasible solutions. Existing CPDP methods assume that the available source data are well labeled. However, due to expensive human efforts for labeling a large amount of defect data, usually, we can only utilize the suitable unlabeled source data. We call CPDP in this scenario as cross-project semisupervised defect prediction (CSDP). Although some within-project semisupervised defect prediction (WSDP) methods have been developed in recent years, there still exists much room for improvement on prediction performance. In this paper, we aim to provide a unified and effective solution for both CSDP and WSDP problems. We introduce the semisupervised dictionary learning technique and propose a cost-sensitive kernelized semisupervised dictionary learning (CKSDL) approach. CKSDL can make full use of the limited labeled defect data and a large amount of unlabeled data in the kernel space. In addition, CKSDL considers the misclassification costs in the dictionary learning process. Extensive experiments on 16 projects indicate that CKSDL outperforms state-of-the-art WSDP methods, using unlabeled cross-project defect data can help improve the WSDP performance, and CKSDL generally obtains significantly better prediction performance than related SSDP methods in the CSDP scenario.

With the ubiquitous deployment of the mobile devices and the explosive growth of Internet traffic, an emerging method called association rules mining (ARM) is proposed to solve the problem of mining potential value of existing big data. However, massive ARM methods focus on positive rules which are easy to ignore interesting information because of negative ones. This paper studies a practical problem of combing negative rules in ARM research. Specifically, we propose a rule certainty updating strategy (RCUS) to combine positive rules with negative rules, which consists of two parts: initialization and updating. To solve the large scale problem with negative rules, the proposed strategy decomposes the large scale problem into several relatively small ones by an improved multiobjective evolutionary algorithm (MOEA) with gene representation and certainty. Simulation results show that our method is outstanding when the scale of attributes and examples is increasing.

According to the practical requirements of high recovery precision and low latency in wireless big sensory data networks, this paper proposes an accelerated distributed rate control method for minimizing the recovery error of big sensory data. This method can guarantee the error minimization of reconstructed data and converge to the optimal value fast with a lower latency. In order to achieve these effects, an accelerated distributed solving algorithm is constructed by designing accelerated subgradient method for dual decomposition. This solving algorithm achieves convergence rate O(1/t2) in practical implementation, which significantly improves the convergence rate of regular solving algorithms. Meanwhile, the convergence analysis testifies the convergence property of the proposed distributed solving algorithm, and this algorithm is applicable to other convex optimization problems. Finally, the performance evaluation shows that the proposed accelerated method can converge to the unique optimal value successfully and the convergence speed is faster than the regular optimization method, and this proposed method can be extended to networks of different sizes without sacrificing the accelerated effect.

The advances of network technology and mobile communication technology are making eHealth possible. In eHealth systems, physiological data and relevant context-aware data are acquired continuously and in real time. At the same time, such large-scale data results in huge challenges in the aspect of real-time big data processing since eHealth data appears in the form of data stream. Therefore, we propose a novel incremental learning algorithm, namely α-SVMSGD, which improves the SVMSGD (Support Vector Machine-Stochastic Gradient Descent) algorithm by updating the training data with the continuous data stream. Besides, this α-SVMSGD may handle the problem that original SVMSGD cannot further mine the useful information in unclassified data. In α-SVMSGD, the process of training data updating is completed by introducing the concept of forgetting mechanism, in which the forgetting factor α is introduced to weed out useless training data. α-SVMSGD is applied into ambient assisted living communications, and further incorporated into the data filtering layer of a local data processing architecture (LDPA) to reduce data redundancy. Simulation results confirm that the proposed algorithm is a promising data redundancy solution for classification without loss of accuracy in the case of real-time data stream.

Machine-to-Machine (M2M) mobile networks are distributed systems which include various actuators and sensors. In terms of the security of M2M mobile networks, one very significant issue is the security of Sensor Networks (SNs). Particularly, the security of transferring data from sensors to their destinations is very critical. In this paper, focusing on intrusion detection techniques, we propose an attack-defense game model to detect malicious nodes using a repeated game approach. In the proposed game model, attackers and defenders make different strategies to achieve optimal payoffs. The existences of pure nash equilibrium and mixed nash equilibrium are analyzed and proved. In the Intrusion Detection System (IDS), a game tree model is introduced to solve the error detection and missing detection problems. Simulation results present that the proposed model can reduce energy consumption by up to 50% compared with the All Monitor (AM) model, and improve the detection rate by up to 10-15% compared with the Cluster Head (CH) monitor model.

Due to the limited resources of DTMSN (Delay Tolerant Mobile Sensor Networks), network congestion becomes a critical problem to resolve. Traditional congestion control methods where the number of copies is restricted to limit data packet forwarding cannot adapt to constantly changing network environment because of fixed number of copies. Fortunately, this problem can be solved through a real-time algorithm by modifying data packet forwarding conditions. However, one of the major challenges of this algorithm is detecting characteristics of the network environment accurately and efficiently. In this paper, an optimized routing algorithm, RVNS (Reduced Variable Neighborhood Search)-based Spray and Wait (SW) is proposed. In this algorithm, nodes will transmit and store the counter record of each other when they meet, based on which, RVNS is introduced to calculate a real-time threshold for the forwarding condition to control packet delivery. Simulation results show that the proposed algorithm increases delivery probability and dramatically reduces the overhead ratio. In some extreme cases, this algorithm can reach an extremely low overhead ratio (ten times lower than that of SW), meaning that the proposed algorithm suits challenged networks well.

Due to the explosive increment of data in big data era, it is a challenging task to analyze and extract meaningful data for users. Data needs to be timely operated because of the time sensitivity, so it faces enormous pressure in storage and computing. To deal with the problem that it is hard to achieve valuable information from out-of-order streams over big data in short time, a model-matching algorithm based on improved BP (Back Propagation) is proposed. In the algorithm, the matching model is set dynamically. Information is extracted for users according to the order of data's arriving time. Furthermore, the algorithm parameters are automatically adjusted in the process of learning and matching. Accordingly, the responding speed of learning is accelerated and the time of matching reduces. In the simulation, a group of optimum parameters of improved BP are achieved by using self-adapting adjusting mechanism. The threshold (TH), connecting weight (CW) and learning rate (LR) are equal to 1.5, 3 and 1, respectively. We implement our model-matching algorithm on 10000 sets of out-of-order streams with these parameters. Results indicate that the proposed algorithm can obviously improve the accuracy and speed of matching and achieve better stability.

The concept of ring signature was first introduced by Rivest et al. in 2001. In a ring signature, instead of revealing the actual identity of the message signer, it specifies a set of possible signers. The verifier can be convinced that the signature was indeed generated by one of the ring members; however, the verifier is unable to tell which member actually produced the signature. A convertible ring signature scheme allows the real signer to convert a ring signature into an ordinary signature by revealing secret information about the ring signature. Thus, the real signer can prove the ownership of a ring signature if necessary, and the the other members in the ring cannot prove the ownership of a ring signature. Based on the original ElGamal signature scheme, a generalized ring signature scheme was proposed for the first time in 2008. The proposed ring signature can achieve unconditional signer ambiguity and is secure against adaptive chosen-message attack in the random oracle model. By comparing to ring signatures based on RSA algorithm, the authors claimed that the proposed generalized ring signature scheme is convertible. It enables the actual message signer to prove to a verifier that only she is capable of generating the ring signature. Through cryptanalysis, we show that the convertibility of the generalized ring signature scheme cannot be satisfied. Everyone in the ring signature has the ability to claim that she generates the generalized ring signature.