A5016169662- Advanced Graph Neural Networks
- Topic Modeling
- Computational Drug Discovery Methods
- Machine Learning in Materials Science
- Multimodal Machine Learning Applications
- Natural Language Processing Techniques
- Advanced Neural Network Applications
- Analytical Chemistry and Chromatography
- Graph Theory and Algorithms
- Neural Networks and Applications
- Regional Economic and Spatial Analysis
- Machine Learning in Bioinformatics
- Topological and Geometric Data Analysis
- Electrospun Nanofibers in Biomedical Applications
- Drug Solubulity and Delivery Systems
- EEG and Brain-Computer Interfaces
- Silk-based biomaterials and applications
- Caching and Content Delivery
- Functional Brain Connectivity Studies
- Neural dynamics and brain function
- Tensor decomposition and applications
- Intelligent Tutoring Systems and Adaptive Learning
- Carbon and Quantum Dots Applications
- Signaling Pathways in Disease
- Globalization, Economics, and Policies
Guangxi University
2024
Rutgers, The State University of New Jersey
2023-2024
China Institute of Finance and Capital Markets
2024
Tiangong University
2023-2024
Stevens Institute of Technology
2021-2023
China Earthquake Administration
2022
Tokyo University of Science
2020-2021
Georgia Institute of Technology
2017
Emory University
2017
Beihang University
2015-2016
AlphaFold2 revolutionized structural biology with the ability to predict protein structures exceptionally high accuracy. Its implementation, however, lacks code and data required train new models. These are necessary (1) tackle tasks, like protein–ligand complex structure prediction, (2) investigate process by which model learns (3) assess model's capacity generalize unseen regions of fold space. Here we report OpenFold, a fast, memory efficient trainable implementation AlphaFold2. We...
Transformers are considered one of the most important deep learning models since 2018, in part because it establishes state-of-the-art (SOTA) records and could potentially replace existing Deep Neural Networks (DNNs). Despite remarkable triumphs, prolonged turnaround time Transformer is a widely recognized roadblock. The variety sequence lengths imposes additional computing overhead where inputs need to be zero-padded maximum sentence length batch accommodate parallel platforms. This paper...
The expressive power of neural networks in modelling non-trivial distributions can principle be exploited to bypass topological freezing and critical slowing down simulations lattice field theories. Some popular approaches are unable sample correctly topology, which may lead some classes configurations not being generated. In this contribution, we present a novel generative method inspired by model previously introduced the ML community (GFlowNets). We demonstrate its efficiency at exploring...
Large Multimodal Models (LMMs) exhibit major shortfalls when interpreting images and, by some measures, have poorer spatial cognition than small children or animals. Despite this, they attain high scores on many popular visual benchmarks, with headroom rapidly eroded an ongoing surge of model progress. To address there is a pressing need for difficult benchmarks that remain relevant longer. We take this idea to its limit introducing ZeroBench-a lightweight reasoning benchmark entirely...
Six-bit quantization (FP6) can effectively reduce the size of large language models (LLMs) and preserve model quality consistently across varied applications. However, existing systems do not provide Tensor Core support for FP6 struggle to achieve practical performance improvements during LLM inference. It is challenging on GPUs due (1) unfriendly memory access weights with irregular bit-width (2) high runtime overhead weight de-quantization. To address these problems, we propose TC-FPx,...
Shaoyi Huang, Dongkuan Xu, Ian Yen, Yijue Wang, Sung-En Chang, Bingbing Li, Shiyang Chen, Mimi Xie, Sanguthevar Rajasekaran, Hang Liu, Caiwen Ding. Proceedings of the 60th Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2022.
The solubility of a drug is higher when it in an amorphous form than crystalline form. To enhance the ibuprofen (IBU), poorly water-soluble drug, we attempted to adsorb IBU onto spherical porous calcium silicate (Florite® PS300, PS300) two ways: evaporation (EV) and sealed heating (SH) methods. crystallinity samples was evaluated using powder X-ray diffraction analysis (PXRD) differential scanning calorimetry (DSC). molecular interaction between PS300 with FTIR. In addition, dissolution...
Graph learning is becoming increasingly popular due to its superior performance in tackling many grand challenges. While quantization widely used accelerate Neural Network (GNN) computation, quantized training faces remarkable roadblocks. Current GNN systems often experience longer time than their full-precision counterparts for two reasons: (i) addressing the accuracy challenge leads excessive overhead, and (ii) optimization potential exposed by not adequately leveraged. This paper...
Many real-world networks are characterized by being temporal and dynamic, wherein the information signifies changes in connections, such as addition or removal of links between nodes. Employing random walks on these is a crucial technique for understanding structural evolution graphs over time. However, existing state-of-the-art sampling methods designed traditional static graphs, such, they struggle to efficiently handle dynamic aspects networks. This deficiency can be attributed several...
Transformer-based deep learning models have become a ubiquitous vehicle to drive variety of Natural Language Processing (NLP) related tasks beyond their accuracy ceiling. However, these also suffer from two pronounced challenges, that is, gigantic model size and prolonged turnaround time. To this end, we introduce ET. r<u>E</u>-thinks self-attention computation for <u>T</u>ransformer on GPUs with the following contributions: First, novel architecture, which encompasses tailored operators...
This work considers the task of representation learning on attributed relational graph (ARG). Both nodes and edges in an ARG are associated with attributes/features allowing ARGs to encode rich structural information widely observed real applications. Existing neural networks offer limited ability capture complex interactions within local contexts, which hinders them from taking advantage expression power ARGs. We propose motif convolution module (MCM), a new motif-based technique better...
Summary Traditional tracking classification algorithm has been widely applied to target in wireless sensor networks. In this paper, focusing on the accuracy of networks, we propose an improved threshold factor track algorithm. The extracts motion model according intrinsic properties target. It updates iterative center real‐time state moving and timely filters out weak correlated or uncorrelated data. order show is more effective, compare proposed with based Euclidean distance comprehensive...
Molecular similarity search has been widely used in drug discovery to identify structurally similar compounds from large molecular databases rapidly. With the increasing size of chemical libraries, there is growing interest efficient acceleration large-scale search. Existing works mainly focus on CPU and GPU accelerate computation Tanimoto coefficient measuring pairwise between different fingerprints. In this paper, we propose optimize an FPGA-based accelerator design exhaustive approximate...
Conventional wisdom in pruning Transformer-based language models is that reduces the model expressiveness and thus more likely to underfit rather than overfit. However, under trending pretrain-and-finetune paradigm, we postulate a counter-traditional hypothesis, is: increases risk of overfitting when performed at fine-tuning phase. In this paper, aim address problem improve performance via progressive knowledge distillation with error-bound properties. We show for first time reducing can...
Although Transformer-based deep learning models have been widely used in many natural language processing (NLP) tasks as well computer vision, they suffer from gigantic model size and long latency. Network pruning can reduce the computational cost size. However, existing works mainly focus on irregular(sparse) pruning, which often causes irregular computations extra indices per remained weight. In this work, we propose a Tensor-core inspired hierarchical compression method to push...