The prevalence of real-world multi-view data makes incomplete multi-view clustering (IMVC) a crucial research. The rapid development of Graph Neural Networks (GNNs) has established them as one of the mainstream approaches for multi-view clustering. Despite significant progress in GNNs-based IMVC, some challenges remain: (1) Most methods rely on the K-Nearest Neighbors (KNN) algorithm to construct static graphs from raw data, which introduces noise and diminishes the robustness of the graph topology. (2) Existing methods typically utilize the Mean Squared Error (MSE) loss between the reconstructed graph and the sparse adjacency graph directly as the graph reconstruction loss, leading to substantial gradient noise during optimization. To address these issues, we propose a novel Dynamic Deep Graph Learning for Incomplete Multi-View Clustering with Masked Graph Reconstruction Loss (DGIMVCM). Firstly, we construct a missing-robust global graph from the raw data. A graph convolutional embedding layer is then designed to extract primary features and refined dynamic view-specific graph structures, leveraging the global graph for imputation of missing views. This process is complemented by graph structure contrastive learning, which identifies consistency among view-specific graph structures. Secondly, a graph self-attention encoder is introduced to extract high-level representations based on the imputed primary features and view-specific graphs, and is optimized with a masked graph reconstruction loss to mitigate gradient noise during optimization. Finally, a clustering module is constructed and optimized through a pseudo-label self-supervised training mechanism. Extensive experiments on multiple datasets validate the effectiveness and superiority of DGIMVCM.

Tianyu Zong, Rui Dai, Hongzhu Yi, Yuanxiang Wang, Zhenghao Zhang, Zhenyu Guan, Yujia Yang, Bingkang Shi, Yueyang Ding, Xiangxiang Chu, Kaikui Liu, Jungang Xu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Placement is a critical and time-consuming step in very-large-scale integration (VLSI) design flow. As placement methods continue to be researched, they introduce more parameters, making current methods for configuring parameters heavily reliant on human experience for each design. This article proposes a novel cross-design parameter optimization method, iPO, to accelerate parameter tuning without human involvement in different placement engines (like iEDA-iPL and DREAMPlace). Specifically, we introduce a heuristic strategy called Constant Liar to accelerate parameter tuning, allowing us to optimize parameters concurrently on different machines. Our research indicates that optimizing parameters for every design is time-consuming. To address the inefficiency of parameter tuning, we propose a cross-design parameter transfer learning strategy. This strategy measures the cosine similarity between designs in collaboration with a graph embedding algorithm representing netlists and cells. Compared with DREAMPlace on ISPD2015 benchmarks, our method achieves average improvements of 9.8% in half-perimeter wirelength (HPWL) and 12.0% in route congestion. When compared with AutoDMP, iPO shows an average improvement of 11% in HPWL and 12.3% in congestion, along with a 3.49× speed-up in the number of search iterations. Furthermore, we extended our experiments to the iEDA-28nm benchmarks, showing average improvements of 4.7%, 2.7% and 2.8% in HPWL, worst negative slack (WNS) and total negative slack (TNS), respectively, compared with iEDA-iPL. Finally, our ablation studies on parallelization demonstrate that using 10 parallel processes results in approximately an 18× speed-up compared with using a single process.

The increasing complexity of digital circuits and the limitations of heuristic methods have led to growing interest in applying Machine Learning (ML) to Logic Synthesis (LS). ML provides a promising paradigm shift by implementing automated, scalable, and data-driven optimization strategies. This survey provides a comprehensive overview of the latest studies on ML approaches in LS, offering a deep understanding of the fundamental ML methods, and analyzing their strengths and limitations through systematic comparisons. We categorize existing works into two main types: ML-Assistance methods aiming at predicting performance metrics and reducing the cost of traditional simulations, and ML-Agent methods directly replacing heuristic processes in the LS flow. We further analyze ML methods and applications in different LS stages, including Boolean circuit generation, Boolean circuit analysis, logic optimization, and technology mapping, showing great achievements and improvement in exploring non-linear design spaces and discovering new optimization strategies. Finally, we discuss the challenges and limitations in the current situation and further provide a vision of future directions in LS.

Unsupervised sentence embedding representation has become a hot research topic in natural language processing. As a tensor, sentence embedding has two critical properties: direction and norm. Existing works have been limited to constraining only the orientation of the samples' representations while ignoring the features of their module lengths. To address this issue, we propose a new training objective that optimizes the training of unsupervised contrastive learning by constraining the module length features between positive samples. We combine the training objective of Tensor's Norm Constraints with ensemble learning to propose a new Sentence Embedding representation framework, TNCSE. We evaluate seven semantic text similarity tasks, and the results show that TNCSE and derived models are the current state-of-the-art approach; in addition, we conduct extensive zero-shot evaluations, and the results show that TNCSE outperforms other baselines.

Matrix Product States have been extensively explored as a powerful tool for simulating quantum states in image classification task. However, most research has focused on classical simulations or computations involving high-dimensional unitaries, and significant challenges still exist in the practical preparation of Matrix Product States on quantum computers. This paper proposes a novel and practically feasible quantum variational algorithm based on Matrix Product States for image classification task. We design a hardware-efficient quantum circuit with several adjustable entangling operators to prepare the local tensors in Matrix Product States and integrate minimal residuals to ensure computational stability. We demonstrate that our algorithm can reduce the parameter complexity from growing exponentially with the system size to a linear scale. To validate the effectiveness of this quantum variational algorithm, we conducted experiments on the MNIST dataset, achieving accuracies of 99.95% and 95.96% for binary and ten-class classification tasks, which outperforms other related quantum algorithms. This work advances the practical application of quantum machine learning in resource-constrained environments of quantum computing.

As a hybrid quantum-classical algorithm, the variational quantum eigensolver is widely applied in quantum chemistry simulations, especially in computing the electronic structure of complex molecular systems. However, on existing noisy intermediate-scale quantum devices, some factors such as quantum decoherence, measurement errors, and gate operation imprecisions are unavoidable. To overcome these challenges, this study proposes an efficient noise-mitigating variational quantum eigensolver for accurate computation of molecular ground state energies in noisy environments. We design the quantum circuit with reference to the structure of matrix product states and utilize it to pre-train the circuit parameters, which ensures circuit stability and mitigates fluctuations caused by initialization. We also employ zero-noise extrapolation to mitigate quantum noise and combine it with neural networks to improve the accuracy of the noise-fitting function, which significantly eliminates noise interference. Furthermore, we implement an intelligent grouping strategy for measuring Hamiltonian Pauli strings, which not only reduces measurement errors but also improves sampling efficiency. We perform numerical simulations to solve the ground state energy of the H4molecule by using MindSpore Quantum framework, and the results demonstrate that our algorithm can constrain noise errors within the range of $\mathcal{O}\left( {{{10}^{ - 2}}} \right)\sim \mathcal{O}\left( {{{10}^{ - 1}}} \right)$, outperforming mainstream variational quantum eigensolvers. This work provides a new strategy for high-precision quantum chemistry calculations on near-term noisy quantum hardware.

Prompt learning has demonstrated remarkable performance in tuning Vision-Language Models (VLMs) for various downstream tasks. Recent studies have shown the effectiveness of prompt distillation in transferring distribution knowledge between VLM teachers and students. However, existing prompt knowledge distillation methods are limited in diversity, focusing solely on positive probabilities. In this paper, we propose a dual prompt distillation (DPD) method, which teaches the student from both positive and negative aspects. Specifically, during the first phase of teacher training, the positive and negative prompts are both optimized by constructing complementary probability distribution signals. In the second distillation phase, the teacher guides the student with dual prompts - positive prompts to select the correct category and negative prompts to exclude incorrect ones. Extensive experimental results across 11 datasets demonstrate that the proposed DPD method either surpasses or matches the performance of existing state-of-the-art (SOTA) methods in both few-shot learning and domain generalization tasks while maintaining competitive computational efficiency. The corresponding code is available at https://github.com/wdinancy/DPD.

Open-vocabulary object detection has emerged as a prominent trend in modern object detection, aiming to identify objects beyond predefined categories. Early object detection methods typically trained detectors on a set of basic object categories and evaluated them on novel object categories, or required users to input the category label of the object. However, these methods heavily rely on prior knowledge, making them unsuitable for real-world scenarios. In this paper, we propose a LLM-enhanced Open-Vocabulary object Detection framework (LOVD) for practical applications. Our method incorporates a text generation agent and a visual-semantic feature alignment agent, both enhanced by a large language model, and a lightweight YOLO backbone for efficient image feature extraction, which enables the model to achieve both high inference efficiency and strong generalization capabilities. Experimental results on the challenging LVIS dataset demonstrate the effectiveness of our method, outperforming many advanced methods in terms of accuracy and speed.

In addressing the complex challenge of Traffic Signal Control (TSC), Deep Reinforcement Learning (DRL) has emerged as a popular solution. In traditional DRL methods applied to TSC problems, deep neural networks are sensitive to minor input changes, which complicates accurate predictions. This ambiguity hampers algorithm convergence, speed, and overall performance. Additionally, existing DRL methods for TSC employ high-dimensional state spaces, escalating computational complexity. This study addresses these challenges by introducing an innovative approach, SLFMLight, that integrates a stochastic traffic flow model with DRL algorithm for TSC. Our method employs an innovative network update algorithm that integrates traffic flow prediction in Q-value learning process to enhance interpretability and accelerate algorithm convergence. Utilizing mode-based multi-actor networks to handle diverse traffic conditions, SLFMLight excels in decision-making towards complex traffic scenarios, especially in congested ones. Concise state definition improves computational efficiency. SLFMLight contributes to the advancement of intelligent traffic management by providing an effective DRL solution that improves interpretability, efficiency, and adaptability in TSC.

Neural architecture performance predictor is an efficient approach for architecture estimation in Neural Architecture Search (NAS). However, existing predictors based on Graph Neural Networks (GNNs) are deficient in modeling long-range interactions between operation nodes and prone to the problem of over-smoothing, which limits their ability to learn neural architecture representation. Furthermore, some Transformer-based predictors use simple position encodings to improve performance via self-attention mechanism, but they fail to fully exploit the subgraph structure information of the graph. To solve this problem, we propose a novel method to enhance the graph representation of neural architectures by combining GNNs and Transformer blocks. We evaluate the effectiveness of our predictor on NAS-Bench-101 and NAS-bench-201 benchmarks, the discovered architecture on DARTS search space achieves an accuracy of 97.61% on CIFAR-10 dataset, which outperforms traditional position encoding methods such as adjacency and Laplacian matrices. The code of our work is available at \url{https://github.com/GNET}.

Image captioning, as a typical multi-modal task, has received increasing attention and has made significant progress. Recently, the proposal of CLIP has shown strong application potential in various tasks, including image captioning. However, the current approaches simply employ CLIP as a visual feature extractor without fully utilizing its potential benefits from the contrastive learning paradigm. Furthermore, most of them are resource-expensive due to the need for additional large-scale image-text datasets for incremental pre-training or additional pre-trained language models for captioning decoder initialization and finetuning. In this paper, we propose a simple and effective approach that fully utilizes CLIP’s properties and advantages to transfer it to an image captioning model. Specifically, following the obtained visual features, we design a memory retrieval module to introduce more information and design a joint representation module for feature enhancement. Furthermore, our captioning decoder is constructed based on the CLIP text encoder with the reuse of its parameters, which aims to leverage its priori textual embedding knowledge. Experimental results on MSCOCO dataset verify the effectiveness of our proposal.

Most existing works of factual inconsistency detection focus on text summarization of generic articles. In this paper, a clinical trial prediction-based factual inconsistency detection approach is proposed for medical text summarization. Inspired by the fact that medical articles related to a clinical trial can give some evidence to predict the outcome of this trial, we believe that a factual consistent summary of the medical articles can also accurately predict the outcome of the corresponding clinical trial. Therefore, we first gather a novel Clinical Trial Prediction-based summarization (CTPSum) dataset, which is a collection of the outcomes of the clinical trials together with the related medical articles and summaries. We then propose a keyword-aware classification model to predict the outcome (successful/failed) of a clinical trial. If the predicted outcome is correct, the generated summary is considered to be factually consistent with the source document, otherwise, it is labeled as factually inconsistent. To evaluate the proposed approach, we further collect a Fact Inconsistency Detection dataset in the Medical Domain (FIDMD), which includes summaries, medical articles, and binary labels indicating factual consistency or inconsistency. Experimental results demonstrate the effectiveness of the proposed approach, specifically, the clinical trial prediction-based factual inconsistency detection approach outperforms several NLI-based factual inconsistency detection methods on the FIDMD dataset.

Currently, with the development of driving technologies, driverless vehicles gradually are becoming more and more available. Therefore, there would be a long period of time during which self-driving vehicles and human-driven vehicles coexist. However, for a mixed platoon, it is hard to control the formation due to the existence of the manual vehicles resulting in weak robustness and slow consensus rate on this system of platoons because of uncertainties caused by human factors for manual vehicles. In order to solve this problem, we establish models of mixed platoons with mixed types of connected and automated vehicles (CAVs), human-driven vehicles (HDVs) and HDVs without the vehicle awareness device (HDVWs). We subsequently design$\mathcal {H}_{\infty} $controllers for the mixed platoons to realize the formation consensus. In addition, we use the$\mathcal {H}_{\infty} $norm of mixed platoons as the control objective investigating the robustness of the control algorithms in alleviating the platoon uncertainties. Furthermore, conditions are proved to maintain the stability of the mixed platoons, and the stability is analyzed based on the variation of the penetration rate of the manual vehicles. Finally, we formulate conditions for parameters according to the definition of string stability to avoid the collisions of vehicles. The results in this study are tested with simulations and suggest that the presented controllers can ensure the consensus of mixed platoons under uncertainties.

CNN-LSTM based architectures have played an important role in image captioning, but limited by the training efficiency and expression ability, researchers began to explore the CNN-Transformer based models and achieved great success. Meanwhile, almost all recent works adopt Faster R-CNN as the backbone encoder to extract region-level features from given images. However, Faster R-CNN needs a pre-training on an additional dataset, which divides the image captioning task into two stages and limits its potential applications. In this paper, we build a pure Transformer-based model, which integrates image captioning into one stage and realizes end-to-end training. Firstly, we adopt SwinTransformer to replace Faster R-CNN as the backbone encoder to extract grid-level features from given images; Then, referring to Transformer, we build a refining encoder and a decoder. The refining encoder refines the grid features by capturing the intra-relationship between them, and the decoder decodes the refined features into captions word by word. Furthermore, in order to increase the interaction between multi-modal (vision and language) features to enhance the modeling capability, we calculate the mean pooling of grid features as the global feature, then introduce it into refining encoder to refine with grid features together, and add a pre-fusion process of refined global feature and generated words in decoder. To validate the effectiveness of our proposed model, we conduct experiments on MSCOCO dataset. The experimental results compared to existing published works demonstrate that our model achieves new state-of-the-art performances of 138.2% (single model) and 141.0% (ensemble of 4 models) CIDEr scores on 'Karpathy' offline test split and 136.0% (c5) and 138.3% (c40) CIDEr scores on the official online test server. Trained models and source code will be released.

Recent few-shot learning methods based on graph neural networks (GNN) over-focus on the connections between nodes while ignoring the pair-wise relations between nodes. Meta-learning aims at model parameter initialization, enabling the model to gain the generalization capability, which assists GNN to pay more attention to nodes. There are rare methods to apply meta-learning to non-Euclidean spaces (such as graph structures). Thus, we propose a graph meta-learning framework, Meta Edge-labeling Graph Neural Network (Meta-EGNN), to solve image classification in few-shot learning. Meta-EGNN can learn a better parameter initialization for GNN with the prediction of edge labels, which can enhance the generalization of the model on unseen tasks. We also introduce the first-order gradient model-agnostic meta-learning into meta-EGNN, which can not only reduce the computational costs, but also help meta-EGNN extend to the transductive inference. The experimental results on two benchmarks prove that Meta-EGNN is competitive in both supervised and semi-supervised image classification.

Performance estimation of neural architecture is a crucial component of neural architecture search (NAS). Meanwhile, neural predictor is a current mainstream performance estimation method. However, it is a challenging task to train the predictor with few architecture evaluations for efficient NAS. In this paper, we propose a graph masked autoencoder (GMAE) enhanced predictor, which can reduce the dependence on supervision data by self-supervised pre-training with untrained architectures. We compare our GMAE-enhanced predictor with existing predictors in different search spaces, and experimental results show that our predictor has high query utilization. Moreover, GMAE-enhanced predictor with different search strategies can discover competitive architectures in different search spaces. Code and supplementary materials are available at https://github.com/kunjing96/GMAENAS.git.

Graph data is ubiquitous in the real world and graph neural networks (GNNs) are effective for modeling the complex relationships and dependencies between the entities. However, it's difficult to design data-specific GNNs. Recently, researchers have started to apply neural architecture search (NAS) to design GNNs. In this work, we propose a graph architecture search method to decrease the instability with a large number of candidate operations. Following SANE(Search to Aggregate NEighborhood), we focus on searching to aggregate neighbourhoods but we divide the candidate operations into groups. We use a continuous relaxation of our search space and optimize the hyper-networks with a gradient-based algorithm. Extensive experiments on several node-level and graph-level tasks demonstrate that our method achieves a promising performance.

Answer-aware question generation aims to generate answerable questions from a given context and answers. Most of the current models are based on the attention-based sequence-to-sequence ($s$eq2seq) structure. However, these models do not make full use of the answer information, resulting in generating questions unrelated to the answer. We propose an answer driven model, which dynamically incorporates the interactive information between answer and previously generated words in the decoder to help the model decide which aspect of the question to focus on. Further, we use the answer distribution difference as a reward and use reinforcement learning to fine-tune the model. Experimental results show that our model performs better than the baseline models.

Open-domain question answering aims to extract answers from some unlabeled texts. Existing approaches usually follow the retrieve-then-read pipeline, which depends heavily on the quality of the retriever. Therefore, a ranker is significant to denoise those irrelevant paragraphs by giving each candidate paragraph a confidence score. In this paper, we design a light paragraph ranker, which utilizes an iterative approach to combine the paragraph-paragraph relevance and the paragraph-question relevance. Our ranker can fully mine global paragraph information and is lighter than that in most other Open-domain QA models. Furthermore, we explore two types of readers based on whether they sample candidate paragraphs or retain all. Experiments on public datasets show that our model achieves improvements compared to baselines.

In task-oriented dialogue systems, dialogue state tracking (DST) is responsible for estimating the current belief state of a dialogue. Recent research tends to utilize historical information to predict the states. However, most of them lack an efficient attention mechanism to comprehend the utterances well. Besides, existing methods usually ignore the relevance between the current utterances with the earlier ones, which determines whether or not the states change. In this paper, we propose an HDST-MRC (Hierarchical Dialogue State Tracking with Machine Reading Comprehension) model to tackle these issues. Within HSDT-MRC, we introduce bi-directional attention flow to extract a context span as the evidence of the ground truth and leverage another copy-augmented generator to predict the states. Experimental results on MultiWoz 2.0 and MultiWoz 2.1 demonstrate that our model achieves significant improvement compared with the baselines.

Recently, neural architecture search (NAS) has emerged as a technique of growing concern in automatic machine learning (AutoML). Meanwhile, attention-based models, such as attention-based recurrent neural network, transformer-based model, etc., have been widely used in deep learning applications. However, there is no efficient NAS method that can search the architecture of attention-based model so far. To solve this problem, we propose a framework named neural architecture search for attention-based networks (NASABN) by abstracting attention-based models and extracting undefined parts of the model, including the attention layers and cells. NASABN is flexible and general enough to fit different NAS methods, which can also be transferred across different datasets. We conduct extensive experiments with NASABN using gradient descent-based methods like DARTS on Penn Treebank (PTB) and WikiText-2 (WT2) datasets respectively, and achieve competitive performance compared with the state-of-the-art methods.

The Shuffle module is one of the core modules in Spark platform, its performance directly influences the performance and throughput of the whole Spark platform. The existing memory scheduling algorithm for the Shuffle process only equitably allocates tasks according to the number of tasks without considering the different memory requirements of different tasks, which causes memory utilization to drop and low running efficiency when data is skewed. To solve this problem, one self-adaptive memory scheduling algorithm for the Shuffle process (SAMSAS) is proposed in this paper, which does not need to set the priority of task processing in advance. Instead, it can adjust memory allocation self-adaptively through constantly monitoring and learning the actual memory requirements of task execution. The experimental results show that SAMSAS algorithm can improve the utilization rate of the entire memory pool and the running efficiency of each Task, and specially it can effectively improve the running efficiency of Spark platform when processing skew data.

Apache Spark is a fast and general engine for large-scale data processing using distributed memory. It provides different deploy modes to meet the needs of different users and Spark on YARN is the most popular deploy mode. Different deploy modes have different scheduling mechanisms. Spark on YARN has three different schedulers, including FIFO Scheduler, Fair Scheduler, and Capacity Scheduler. However, these three schedulers cannot fit hard real-time application scenarios. With the application of Apache Spark more widely, the needs of hard real-time scheduling will increase quickly. In this paper, we proposed a novel hard real-time scheduling algorithm called DVDA (Deadline and Value Density-Aware) in order to meet the requirements of hard real-time scheduling. Compared with traditional EDF (Earliest Deadline First) algorithm which only considers the deadline, the DVDA algorithm considers both the deadline and value density of the application. Furthermore, we implement a DVDA Scheduler for Spark on YARN based on the DVDA algorithm. Finally, the experiments are conducted to verify the effectiveness of the algorithm. Experimental results show that the proposed algorithm can increase the application completed rate by 18% and 6%, Value Income by 78% and 32% compared with default Capacity scheduler and EDF-Capacity scheduler respectively.

Pseudo relevance feedback (PRF) is commonly used to boost the performance of traditional information retrieval (IR) models by using top-ranked documents to identify and weight new query terms, thereby reducing the effect of query-document vocabulary mismatches.While neural retrieval models have recently demonstrated strong results for adhoc retrieval, combining them with PRF is not straightforward due to incompatibilities between existing PRF approaches and neural architectures.To bridge this gap, we propose an end-to-end neural PRF framework that can be used with existing neural IR models by embedding different neural models as building blocks.Extensive experiments on two standard test collections confirm the effectiveness of the proposed NPRF framework in improving the performance of two state-of-theart neural IR models.

Language models based on Recurrent Neural Networks, e.g. Long Short Term Memory Network (LSTM), have shown strong ability in generating captions from image. However, in previous LSTM-based image captioning models, the image information is input to LSTM at 0th time step, and the network gradually forgets the image information, and only uses the language model to generate a simple description, leaving the potential in generating a better description. To address this challenge, in this paper, a Long-term Recurrent Merge Network (LRMN) model is proposed to merge the image feature at each step via a language model, which not only can improve the accuracy of image captioning, but also can describe the image better. Experimental results show that the proposed LRMN model has a promising improvement in image captioning.

Clinical Decision Support (CDS) can be regarded as an information retrieval (IR) task, where medical records are used to retrieve the full-text biomedical articles to satisfy the information needs from physicians, aiming at better medical solutions. Recent attempts have introduced the advances of deep learning by employing neural IR methods for CDS, where, however, only the document-query relationship is modeled, resulting in non-optimal results in that a medial record can barely reflect the information included in a relevant biomedical article which is usually much longer. Therefore, in addition to the document-query relationship, we propose a document-based neural relevance model (DNRM), addressing the mismatch by utilizing the content of relevant articles to complement the medical records. Specifically, our DNRM model evaluates a document relative to a query and to several pseudo relevant documents for the query at the same time, capturing the interactions from both parts with a feed forward network. Experimental results on the standard Text REtrieval Conference (TREC) CDS track dataset confirm the superior performance of the proposed DNRM model.

With the wide application of Spark big data platform, some problems in practical application are exposed, and one of the main problems is performance optimization. The Shuffle module of Spark is one of the core modules of Spark, and it is also an important module of some other distributed big data computing frameworks. The design of Shuffle module is the key factor that directly determines the performance of big data computing framework. The main optimization parameters of Shuffle process involve the CPU utilization, I/O literacy rate, network transmission rate, and one of these factors is likely to be the bottleneck during the execution of application. The network data transmission time consumption, I/O read and write time, and the CPU utilization are closely related with the size of the data processing. As a result, Spark provides compression configuration options and different compression algorithms for users to select. Different compression algorithms have different effects in compression rate and compression ratio, but the default configuration is usually selected by all users even though they run different applications, so the optimal configuration cannot be achieved. In order to achieve the optimal configuration of compression algorithm for the Shuffle process, one cost optimization model for Spark Shuffle process is proposed in this paper, which enables users to get the best compression configuration before application execution. The experimental results show that the prediction model for compression configuration has an accuracy of 58.3%, and the proposed cost optimization model can improve the performance by 48.9%.

Data representation is a fundamental task in machine learning, which affects the performance of the whole machine learning system. In the past few years, with the rapid development of deep learning, the models for word embedding based on neural networks have brought new inspiration to the research of natural language processing. In this paper, two kinds of schemes for improving the Continuous Bag-of-Words (CBOW) model are proposed. On one hand, the relative positions of adjacent words are taken as weights for the input layer of the model; on the other hand, the context is considered, and which can take part in the training course when the prediction of next target word is to be made. Experimental results show that our proposed models outperform the classical CBOW model.

Clinical Decision Support (CDS) is widely seen as an information retrieval (IR) application in the medical domain. The goal of CDS is to help physicians find useful information from a collection of medical articles with respect to the given patient records, in order to take the best care of their patients. Most of the existing CDS methods do not sufficiently consider the semantic relation between texts, hence the potential in improving the performance in biomedical articles retrieval. This paper proposes a novel feedback-based approach which considers the semantic association between a retrieved biomedical article and a pseudo feedback set. Evaluation results show that our method outperforms the strong baselines and is able to improve over the best runs in the TREC CDS tasks.

In recent years, the development of Internet enables the rapid growth of global data volume, the arrival of the era of big data has brought great challenges to the traditional computing. Big Data systems, such as hadoop, spark, are becoming important platforms to handle big data, but due to design flaws of big data application itself, and unreasonable distributed framework configuration, the performance of the applications in big data system is difficult to achieve peak speed of computer theory, so how to locate performance bottleneck of big data system and analyze the bottleneck causes is worthy of research. In this paper, a 5-layer performance evaluation model for big data system is proposed, which is a reliable basis for performance analysis, and at the same time, a performance optimization model for big data system is also proposed, which can assist performance bottleneck location and bottleneck analysis, and further optimize performance. Based on these two performance models, an event-based performance tool to profile performance data is implemented. Experimental results show that these two performance models are effective for performance evaluation and optimization of big data system, which can improve average running time of big data system by 19%.

Learning to rank plays a very important role in information retrieval. Existing works mainly focus on applying one ranking model to all samples, which may not be suitable for the reality. In this paper, a new method for learning to rank based on query-level vector extraction is proposed, in which we assume that all samples can be divided into multiple parts, and each part is used to train one set of parameters for the model. Based on this assumption, we extracted query-level vector and proposed a dataset partition method based on k-means++, which is used to optimize the ListNet method and RankNet method. Experimental results show that our assumption is right and our method plays a very important role in improving the performance of ListNet and RankNet, and which is also easy to be extended to other learning to rank methods.

Automated essay-scoring (AES) systems utilize computer techniques and algorithms to automatically rate essays written in an educational setting, by which the workload of human raters is greatly reduced. AES is usually addressed as a classification or regression problem, where classical machine learning algorithms such as K-nearest neighbor and support vector machines are applied. In this paper, we argue that essay rating is based on the comparison of writing quality between essays and treat AES rather as a ranking problem by capturing the difference in writing quality between essays. We propose a rank-based approach that trains an essay-rating model by learning to rank algorithms, which have been widely used in many information retrieval and social Web mining tasks. Various linguistic and statistical features are utilized to facilitate the learning algorithms. Extensive experiments on two public English essay datasets, Automated Student Assessment Prize and Chinese Learners English Corpus, show that our proposed approach based on pairwise learning outperforms previous classification or regression-based methods on all 15 topics. Finally, analysis on the importance of the features extracted reveals that content, organization and structure are the main factors that affect the ratings of essays written by native English speakers, while non-native speakers are prone to losing ratings on improper term usage, syntactic complexity and grammar errors.

Nowadays, private clouds are widely used for resource sharing. Hadoop-based clusters are the most popular implementations for private clouds. However, because workload traces are not publicly available, few previous work compares and evaluates different cloud solutions with publicly available benchmarks. In this paper, we use a recently-released Cloud benchmarks suite - CloudRank-D to quantitatively evaluate five different Hadoop task schedulers, including FIFO, capacity, naïve fair sharing, fair sharing with delay, and HOD (Hadoop On Demand) scheduling. Our experiments show that with an appropriate scheduler, the throughput of a private cloud can be improved by 20%.

Transductive learning is a semi-supervised learning paradigm that can leverage unlabeled data by creating pseudo labels for learning a ranking model, when there is only limited or no training examples available. However, the effectiveness of transductive learning in information retrieval (IR) can be hindered by the low quality pseudo labels. To this end, we propose to incorporate a two-step k-means clustering algorithm to select the high quality training queries for generating the pseudo labels. In particular, the first step selects the high-quality queries for which the relevant documents are highly coherent as indicated by the clustering results. The second step then selects the initial training examples for the transductive learning that iteratively aggregating the pseudo examples. Finally, the learning to rank (LTR) algorithms are applied to learn the ranking model using the pseudo training examples created by the transductive learning process. Our proposed approach is particularly suitable for applications where there is only little or no human labels available as it does not necessarily involve the use of relevance assessments information or human efforts. Experimental results on the standard TREC Tweets11 collection show that our proposed approach outperforms strong baselines, namely the conventional applications of learning to rank algorithms using human labels for the training and transductive learning using all the queries available.

Bayesian Network parameter learning is one of the core issues of Bayesian Network research. The parameter estimation of Bayesian Network from large incomplete dataset can be very compute-intensive. A factor graph based Bayesian Network parameter learning algorithm using MapReduce is presented in this paper, which decomposes one Bayesian Network into factors and gets the Bayesian Network parameter through computing the conditional probability tables of each factor independently using Expectation Maximization (EM) algorithm within MapReduce framework. Experimental results show that when the number of training samples is 107, the speed of this parallel algorithm can get 2~6 times the speed of Sequential Expectation Maximization. The algorithm can reduce the training time significantly with increasing the number of Hadoop nodes. Compared with the existing parallel EM method using MapReduce, this algorithm has also a higher speed and can avoid the problem of load imbalance at the same time.