A5101638214- Advanced Neural Network Applications
- Parallel Computing and Optimization Techniques
- Topic Modeling
- Natural Language Processing Techniques
- Functional Brain Connectivity Studies
- Advanced Graph Neural Networks
- Neural Networks and Applications
- Multimodal Machine Learning Applications
- Neural dynamics and brain function
- Brain Tumor Detection and Classification
- Stochastic Gradient Optimization Techniques
- Speech and dialogue systems
- Graph Theory and Algorithms
- Adversarial Robustness in Machine Learning
- Data Management and Algorithms
- Advanced MRI Techniques and Applications
- Tensor decomposition and applications
- Distributed and Parallel Computing Systems
- Health, Environment, Cognitive Aging
- Advanced Vision and Imaging
- Advanced Data Storage Technologies
- Machine Learning and Data Classification
- Land Use and Ecosystem Services
- Memory and Neural Mechanisms
- Seismic Imaging and Inversion Techniques
Tianjin University
2025
National Clinical Research
2025
Capital Medical University
2024-2025
Amazon (United States)
2019-2024
Capital Normal University
2023-2024
Binzhou University
2023-2024
Hangzhou Dianzi University
2008-2024
Beijing Anding Hospital
2024
Shanghai Jiao Tong University
2023
Sungkyunkwan University
2023
The rate of progress in human neurosciences is limited by the inability to easily apply a wide range analysis methods plethora different datasets acquired labs around world. In this work, we introduce framework for creating, testing, versioning and archiving portable applications analyzing neuroimaging data organized described compliance with Brain Imaging Data Structure (BIDS). portability these (BIDS Apps) achieved using container technologies that encapsulate all binary other dependencies...
Abstract Intrinsic disorder (ID) in proteins is well-established structural biology, with increasing evidence for its involvement essential biological processes. As measuring dynamic ID behavior experimentally on a large scale remains difficult, scores of published predictors have tried to fill this gap. Unfortunately, their heterogeneity makes it difficult compare performance, confounding biologists wanting make an informed choice. To address issue, the Critical Assessment protein Disorder...
Functional magnetic resonance imaging (fMRI) offers a rich source of data for studying the neural basis cognition. Here, we describe Brain Imaging Analysis Kit (BrainIAK), an open-source, free Python package that provides computationally optimized solutions to key problems in advanced fMRI analysis. A variety techniques are presently included BrainIAK: intersubject correlation (ISC) and functional connectivity (ISFC), alignment via shared response model (SRM), full matrix analysis (FCMA),...
Large deep learning models have recently garnered substantial attention from both academia and industry. Nonetheless, frequent failures are observed during large model training due to large-scale resources involved extended time. Existing solutions significant failure recovery costs the severe restriction imposed by bandwidth of remote storage in which they store checkpoints.
We study in this paper the Web forum crawling problem, which is a very fundamental step many applications, such as search engine and data mining. As typical user-created content (UCC), has become an important resource on due to its rich information contributed by millions of Internet users every day. However, not trivial problem in-depth link structures, large amount duplicate pages, well invalid pages caused login failure issues. In paper, we propose build prototype intelligent crawler,...
The popularity of Convolutional Neural Network (CNN) models and the ubiquity CPUs imply that better performance CNN model inference on can deliver significant gain to a large number users. To improve CPUs, current approaches like MXNet Intel OpenVINO usually treat as graph use high-performance libraries such MKL-DNN implement operations graph. While achieving reasonable individual from off-the-shelf libraries, this solution makes it inflexible conduct optimizations at level, local...
Recently there has been a surge of research on improving the communication efficiency distributed training. However, little work done to systematically understand whether network is bottleneck and what extent.
Graph neural networks (GNNs) are gaining popularity as a promising approach to machine learning on graphs. Unlike traditional graph workloads where each vertex/edge is associated with scalar, GNNs attach feature tensor vertex/edge. This additional dimension, along consequently more complex vertex- and edge-wise computations, has enormous implications locality parallelism, which existing processing systems fail exploit. paper proposes FeatGraph accelerate GNN by co-optimizing traversal...
As deep learning models nowadays are widely adopted by both cloud services and edge devices, reducing the latency of model inferences becomes crucial to provide efficient serving. However, it is challenging develop tensor programs for operators due high complexity modern accelerators rapidly growing number operators. Deep compilers, such as Apache TVM, adopt declarative scheduling primitives lower bar developing programs. we show that this approach insufficient cover state-of-the-art program...
<sec> <title>BACKGROUND</title> Late-life depression (LLD) is a growing public health concern, particularly in aging populations like China, where traditional treatments often fail to meet elderly patients' unique needs. While digital mindfulness interventions show promise for anxiety, depression, and insomnia, their efficacy neuroelectric mechanisms remain underexplored. This study investigates the FocusZen Mindfulness Stress Reduction System, novel intervention integrating real-time EEG...
<title>Abstract</title> <bold>Background:</bold> Limited research has explored the associations between sleep disturbances (SD) and cognitive impairment (CI) in elderly patients with depression, particularly by incorporating polysomnography (PSG) to assess quality. This study was conducted determine correlations PSG-quantified parameters CI among individuals late-life depression. <bold>Methods:</bold> 65 depression were included study. The status assessed using PSG, while function evaluated...
Web forums have become an important data resource for many web applications, but extracting structured from unstructured forum pages is still a challenging task due to both complex page layout designs and unrestricted user created posts. In this paper, we study the problem of extraction various sites. Our target find solution as general possible extract data, such post title, author, time, content any site. contrast most existing information methods, which only leverage knowledge inside...
Modern deep learning applications urge to push the model inference taking place at edge devices for multiple reasons such as achieving shorter latency, relieving burden of network connecting cloud, and protecting user privacy. The Convolutional Neural Network (CNN) is one most widely used family in applications. Given high computational complexity CNN models, it favorable execute them on integrated GPUs devices, which are ubiquitous have more power better energy efficiency than accompanying...
Functional magnetic resonance imaging (fMRI) offers a rich source of data for studying the neural basis cognition. Here, we describe Brain Imaging Analysis Kit (BrainIAK), an open-source, free Python package that provides computationally-optimized solutions to key problems in advanced fMRI analysis. A variety techniques are presently included BrainIAK: intersubject correlation (ISC) and functional connectivity (ISFC), alignment via shared response model (SRM), full matrix analysis (FCMA),...
Although pre-trained language models have remarkably enhanced the generation ability of dialogue systems, open-domain Chinese systems are still limited by data and model size compared with English ones. In this paper, we propose EVA, a system that contains largest 2.8B parameters. To build model, collect dataset named WDC-Dialogue from various public social media. This 1.4B context-response pairs is used as pre-training corpus EVA. Extensive experiments on automatic human evaluation show EVA...
Multi-task model training has been adopted to enable a single deep neural network (often large language model) handle multiple tasks (e.g., question answering and text summarization). commonly receives input sequences of highly different lengths due the diverse contexts tasks. Padding (to same sequence length) or packing (short examples into long is usually prepare samples for training, which nonetheless not space computation efficient. This paper proposes dynamic micro-batching approach...
Yida Wang, Yinhe Zheng, Yong Jiang, Minlie Huang. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.