A5002016490- Advanced Graph Neural Networks
- Graph Theory and Algorithms
- Machine Learning in Materials Science
- Ferroelectric and Negative Capacitance Devices
- Complex Network Analysis Techniques
- Computational Drug Discovery Methods
- Recommender Systems and Techniques
- Topic Modeling
- Liver physiology and pathology
- Parallel Computing and Optimization Techniques
- Biofield Effects and Biophysics
- Nutrition and Health in Aging
- Schizophrenia research and treatment
- Advanced Memory and Neural Computing
- Cancer-related gene regulation
- Cancer Research and Treatments
- Adversarial Robustness in Machine Learning
- Psychosomatic Disorders and Their Treatments
Shanghai Jiao Tong University
2022-2024
Shanghai Mental Health Center
2024
Ruijin Hospital
2022
Shanghai Artificial Intelligence Laboratory
2020-2021
Amazon (Germany)
2021
Children's Hospital of Chongqing Medical University
2012
Chongqing Medical University
2012
Advancing research in the emerging field of deep graph learning requires new tools to support tensor computation over graphs. In this paper, we present design principles and implementation Deep Graph Library (DGL). DGL distills computational patterns GNNs into a few generalized sparse operations suitable for extensive parallelization. By advocating as central programming abstraction, can perform optimizations transparently. cautiously adopting framework-neutral design, allows users easily...
Graph neural networks (GNNs) constitute a class of deep learning methods for graph data. They have wide applications in chemistry and biology, such as molecular property prediction, reaction drug-target interaction prediction. Despite the interest, GNN-based modeling is challenging it requires data preprocessing addition to programming learning. Here, we present Deep Library (DGL)-LifeSci, an open-source package on graphs life science. (DGL)-LifeSci python toolkit based RDKit, PyTorch,...
Graph neural networks (GNN) have shown great success in learning from graph-structured data. They are widely used various applications, such as recommendation, fraud detection, and search. In these domains, the graphs typically large, containing hundreds of millions nodes several billions edges. To tackle this challenge, we develop DistDGL, a system for training GNNs mini-batch fashion on cluster machines. DistDGL is based Deep Library (DGL), popular GNN development framework. distributes...
While many systems have been developed to train graph neural networks (GNNs), efficient model evaluation, which computes node embedding according a given model, remains be addressed. For instance, using the widely adopted node-wise approach, evaluation can account for over 90% of time in end-to-end training process due neighbor explosion, means that accesses its multi-hop neighbors. The layer-wise approach avoids explosion by conducting computation layer GNN models. However, takes...
Graph neural networks (GNN) have shown great success in learning from graph-structured data. They are widely used various applications, such as recommendation, fraud detection, and search. In these domains, the graphs typically large, containing hundreds of millions nodes several billions edges. To tackle this challenge, we develop DistDGL, a system for training GNNs mini-batch fashion on cluster machines. DistDGL is based Deep Library (DGL), popular GNN development framework. distributes...
Despite the recent success of graph neural networks (GNN), common architectures often exhibit significant limitations, including sensitivity to oversmoothing, long-range dependencies, and spurious edges, e.g., as can occur a result heterophily or adversarial attacks. To at least partially address these issues within simple transparent framework, we consider new family GNN layers designed mimic integrate update rules two classical iterative algorithms, namely, proximal gradient descent...
Cognitive impairment is common in chronic schizophrenia patients. The purpose of this study was to explore the efficacy Shen-based Qigong Exercise (SBQE) improving cognitive stable patients rehabilitation wards. SBQE derived from theory "body-spirit syncretism (xin shen he yi)" traditional Chinese medicine (TCM) and extracted four techniques. In 12-week, randomized, single-blind, controlled study, a total 40 were randomly assigned either group or control group. scores for Scale Assessment...
Sarcopenia and cognitive impairment are the most prevalent causes of disability in older individuals. The aim this study was to assess prevalence sarcopenia association between patients.A cross-sectional undertaken, comprised 250 male patients aged 65 over. defined using diagnostic recommended consensus by Asian Working Group for sarcopenia, participants were classified into non-sarcopenia groups according definition. functions assessed Mini-Mental State Examination (MMSE). After bivariate...
While many systems have been developed to train Graph Neural Networks (GNNs), efficient model inference and evaluation remain be addressed. For instance, using the widely adopted node-wise approach, can account for up 94% of time in end-to-end training process due neighbor explosion, which means that a node accesses its multi-hop neighbors. On other hand, layer-wise avoids explosion problem by conducting layer such nodes only need their one-hop neighbors each layer. However, implementing...