A5100328838- Topic Modeling
- Natural Language Processing Techniques
- Music and Audio Processing
- Speech Recognition and Synthesis
- Speech and Audio Processing
- Advanced Graph Neural Networks
- Time Series Analysis and Forecasting
- Complex Network Analysis Techniques
- Machine Learning in Healthcare
- Privacy-Preserving Technologies in Data
- Domain Adaptation and Few-Shot Learning
- Advanced Computational Techniques and Applications
- Stochastic Gradient Optimization Techniques
- Neural Networks and Applications
- Energy Load and Power Forecasting
- Biomedical Text Mining and Ontologies
- Network Security and Intrusion Detection
- Stock Market Forecasting Methods
- Internet Traffic Analysis and Secure E-voting
- Artificial Intelligence in Healthcare
- Anomaly Detection Techniques and Applications
- Digital Media Forensic Detection
- Interconnection Networks and Systems
- Mobile Crowdsensing and Crowdsourcing
- Age of Information Optimization
Liaoning Normal University
2024
Tsinghua University
2024
PLA Information Engineering University
2024
North China University of Technology
2024
Zhengzhou University
2023
Beihang University
2017-2023
Guangzhou University
2023
ETH Zurich
2021-2023
L'Oreal (United States)
2023
Chinese Academy of Sciences
2020-2022
Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long time-series forecasting (LSTF) demands a high capacity model, which is ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown potential Transformer increase capacity. However, there are several severe issues with that prevent it from being directly applicable LSTF, including quadratic time...
Audio deepfake detection is an emerging topic, which was included in the ASVspoof 2021. However, recent shared tasks have not covered many real-life and challenging scenarios. The first Deep synthesis Detection challenge (ADD) motivated to fill gap. ADD 2022 includes three tracks: low-quality fake audio (LF), partially (PF) game (FG). LF track focuses on dealing with bona fide fully utterances various real-world noises etc. PF aims distinguish from real. FG a rivalry game, two tasks:...
Named entity recognition (NER) is a widely studied task in natural language processing. Recently, growing number of studies have focused on the nested NER. The span-based methods, considering as span classification task, can deal with entities naturally. But they suffer from huge search space and lack interactions between entities. To address these issues, we propose novel sequence-to-set neural network for Instead specifying candidate spans advance, provide fixed set learnable vectors to...
Graph neural networks (GNNs) have shown excellent performance in a wide range of applications such as recommendation, risk control, and drug discovery. With the increase volume graph data, distributed GNN systems become essential to support efficient training. However, existing training suffer from various issues including high network communication cost, low CPU utilization, poor end-to-end performance. In this paper, we propose ByteGNN, which addresses limitations with three key designs:...
Named entity recognition (NER) is a fundamental task to recognize specific types of entities from given sentence. Depending on how the appear in sentence, it can be divided into three subtasks, namely, Flat NER, Nested and Discontinuous NER. Among existing approaches, only generative model uniformly adapted these subtasks. However, when applied its optimization objective not consistent with task, which makes vulnerable incorrect biases. In this paper, we analyze biases generation process...
Many real-world applications require the prediction of long sequence time-series, such as electricity consumption planning. Long time-series forecasting (LSTF) demands a high capacity model, which is ability to capture precise long-range dependency coupling between output and input efficiently. Recent studies have shown potential Transformer increase capacity. However, there are several severe issues with that prevent it from being directly applicable LSTF, including quadratic time...
Anomalous subgraph detection has been successfully applied to event in social media. However, the problembecomes challenging when media network incorporates abundant attributes, which leads a multivariate network. The characteristic makes most existing methods incapable tackle this problem effectively and efficiently, as it involves joint feature selection that not well addressed current literature, especially, dynamic networks attributes evolve over time.
At online retail platforms, it is crucial to actively detect the risks of transactions improve customer experience and minimize financial loss. In this work, we propose xFraud, an explainable fraud transaction prediction framework which mainly composed a detector explainer. The xFraud can effectively efficiently predict legitimacy incoming transactions. Specifically, utilizes heterogeneous graph neural network learn expressive representations from informative heterogeneously typed entities...
The traffic dynamics of multi-layer networks has become a hot research topic since many are comprised two or more layers subnetworks. Due to its low capacity, the traditional shortest path routing (SPR) protocol is susceptible congestion on two-layer complex networks. In this paper, we propose an efficient strategy named improved global awareness (IGAR) which based betweenness centrality nodes in layers. With proposed strategy, paths can bypass hub both enhance transport efficiency....
Mining from graph-structured data is an integral component of graph management. A recent trending technique, convolutional network (GCN), has gained momentum in the mining field, and plays essential part numerous graph-related tasks. Although emerging GCN optimization techniques bring improvements to specific scenarios, they perform diversely different applications introduce many trial-and-error costs for practitioners. Moreover, existing models often suffer oversmoothing problem. Besides,...
The transport efficiency of a network is strongly related to the underlying structure. In this paper, we propose an efficient strategy named high-betweenness-first (HBF) for purpose improving traffic handling capacity scale-free networks by limiting fraction undirected links be unidirectional ones based on links’ betweenness. Compared with high-degree-first (HDF) strategy, can more significantly enhanced under proposed link-directed shortest path (SP) routing protocol. Simulation results in...
Lijie Wang, Yaozong Shen, Shuyuan Peng, Shuai Zhang, Xinyan Xiao, Hao Liu, Hongxuan Tang, Ying Chen, Hua Wu, Haifeng Wang. Proceedings of the 26th Conference on Computational Natural Language Learning (CoNLL). 2022.
Recently, numerous efforts have continued to push up performance boundaries of document-level relation extraction (DocRE) and claimed significant progress in DocRE. In this paper, we do not aim at proposing a novel model for Instead, take closer look the field see if these gains are actually true. By taking comprehensive literature review thorough examination popular DocRE datasets, find that achieved upon strong or even untenable assumption common: all named entities perfectly localized,...
Federated learning (FL) is an emerging distributed machine method that empowers in-situ model training on decentralized edge devices. However, multiple simultaneous FL tasks could overload resource-constrained In this work, we propose the first system to effectively coordinate and train tasks. We formalize problem of Then, present our new approach, MAS (Merge Split), optimize performance starts by merging into all-in-one task with a multi-task architecture. After for few rounds, splits two...
Traffic capacity is critical for various networks and strongly depends on the distribution of link's bandwidth resources. In this paper, we propose a betweenness-based allocation strategy in which each link l ij allocated proportionally to product (1 + B i ) α j , where tunable parameter, are betweenness node j, respectively. The optimal value achieved by extensive simulations slightly increases with network size. Our new achieves highest traffic when compared average previously proposed...
Reasoning is a fundamental problem for computers and deeply studied in Artificial Intelligence. In this paper, we specifically focus on answering multi-hop logical queries Knowledge Graphs (KGs). This complicated task because, real-world scenarios, the graphs tend to be large incomplete. Most previous works have been unable create models that accept full First-Order Logical (FOL) queries, which include negative only able process limited set of query structures. Additionally, most methods...
Transducer-based models, such as RNN-Transducer and transformer-transducer, have achieved great success in speech recognition. A typical transducer model decodes the output sequence conditioned on current acoustic state previously predicted tokens step by step. Statistically, The number of blank prediction results accounts for nearly 90\% all tokens. It takes a lot computation time to predict tokens, but only non-blank will appear final sequence. Therefore, we propose method named fast-skip...
Molecular Relational Learning (MRL), aiming to understand interactions between molecular pairs, plays a pivotal role in advancing biochemical research. Recently, the adoption of large language models (LLMs), known for their vast knowledge repositories and advanced logical inference capabilities, has emerged as promising way efficient effective MRL. Despite potential, these methods predominantly rely on textual data, thus not fully harnessing wealth structural information inherent graphs....
As The Onion Router (Tor) becomes increasingly prevalent, attackers have initiated Sybil attacks by controlling a plethora of malicious relay nodes, severely compromising user privacy. Thus, the identification nodes is crucial for ensuring security Tor network. Previous studies suggested that tend to similar configuration and close uptime, leading design Nearest-neighbor ranking algorithm analyzing similarity between nodes. This was followed further manual analysis sift from ones. However,...
Abstract The search of the new physics (NP) beyond Standard Model is one most important topics in current high energy physics. With increasing luminosities at colliders, for NP signals requires analysis more and data, efficiency data processing becomes particularly important. As a machine learning algorithm, support vector (SVM) expected to be useful NP. Meanwhile, quantum computing has potential offer huge advantages when dealing with large amounts which suggests that SVM (QSVM) tool future...
With the rapid development of information technologies, which facilitates perfection healthcare systems, a variety clinical data is becoming available. The patient Electronic Health Records (EHR) one important sources in on conducts personalized medicine. However, it challenging if raw EHRs are directly used to conduct related medical prediction due its heterogeneity, sparsity and existence noise. To address this issue, paper proposes an integrative driven approach called Medical Temporal...