- Protein Structure and Dynamics
- Complex Network Analysis Techniques
- Natural Language Processing Techniques
- Topic Modeling
- Evolutionary Algorithms and Applications
- RNA and protein synthesis mechanisms
- Software Engineering Research
- Gene Regulatory Network Analysis
- Computational Drug Discovery Methods
- Discourse Analysis and Cultural Communication
- Evolution and Genetic Dynamics
- Domain Adaptation and Few-Shot Learning
- Bioinformatics and Genomic Networks
- Advanced Neural Network Applications
- Advanced Graph Neural Networks
- Computational and Text Analysis Methods
- Neuroscience and Neural Engineering
- Machine Learning in Materials Science
- Adversarial Robustness in Machine Learning
- Wave and Wind Energy Systems
- Image and Video Quality Assessment
- Ocean Waves and Remote Sensing
- Advanced Malware Detection Techniques
- Statistical Distribution Estimation and Applications
- Multimodal Machine Learning Applications
Tsinghua University
2023-2024
Michigan State University
2020-2024
Beihang University
2021-2024
Chongqing University
2024
Civil Aviation Flight University of China
2024
Yunnan University
2018-2022
Nanyang Technological University
2022
Carnegie Mellon University
2021
Shandong University of Science and Technology
2021
Soochow University
2020
Significant research progress has been made in the field of protein structure and fitness prediction. Particularly, single-sequence-based prediction methods like ESMFold OmegaFold achieve a balance between inference speed accuracy, showing promise for many downstream tasks. Here, we propose SPIRED, model that exhibits comparable performance to state-of-the-art but with approximately 5-fold acceleration at least one order magnitude reduction training consumption. By integrating SPIRED neural...
The industrial 4.0 era is the fourth revolution and characterized by network penetration; therefore, traditional manufacturing value creation will undergo revolutionary changes. Artificial intelligence drive next technology revolution, knowledge graphs comprise main foundation of this revolution. intellectualization information an important part industry 4.0, we can efficiently integrate multisource heterogeneous data realize through powerful semantic association graphs. Knowledge have been...
Binary code learning, also known as hashing, has received increasing attention in large-scale visual search. By transforming high-dimensional features to binary codes, the original Euclidean distance is approximated via Hamming distance. More recently, it advocated that manifold distance, rather than should be preserved space. However, retains an open problem directly preserve structure by hashing. In particular, first needs build local linear embedding feature space, and then quantize such...
A bstract Significant research progress has been made in the field of protein structure and fitness prediction. Particularly, single-sequence-based prediction methods like ESMFold OmegaFold achieve a balance between inference speed accuracy, showing promise for many downstream tasks. Here, we propose SPIRED, novel model that exhibits comparable performance to state-of-the-art but with approximately 5-fold acceleration at least one order magnitude reduction training consumption. By...
A brain tumor is a growth of abnormal cells in the tissues brain, which difficult for treatment and severely affects patients' cognitive ability. Recent year magnetic resonance imaging (MRI) has been widely used technique to assess tumors. However manual segmentation artificial extracting features block MRI's practice when facing with huge amount data produced by MRI. An efficient automatic image still needed. In this paper, novel framework tumors, have 5 parts resnet-50 use as backbone,...
Crowdsourced testing has been widely used to improve software quality as it can detect various bugs and simulate real usage scenarios. workers perform tasks on crowdsourcing platforms present their experiences test reports, which naturally generates an overwhelming number of reports. Therefore, inspecting these reports becomes a time-consuming yet inevitable task. In recent years, many text-based prioritization clustering techniques have proposed address this challenge. However, in mobile...
A bstract Accurate prediction of the fitness and stability a protein upon mutations is high importance in engineering design. Despite rapid development deep learning techniques accumulation experimental data, multi-labeled nature data hinders training robust deep-learning-based models for tasks. Here, we propose three geometric-learning-based models, GeoFitness, GeoDDG GeoDTm, score, ΔΔ G Δ T m mutations, respectively. In optimization designed novel loss function to allow supervised unified...
<title>Abstract</title> Accurate prediction of the fitness and stability a protein upon mutations is high importance in engineering design. Despite rapid development deep learning techniques accumulation experimental data, multi-labeled nature data hinders training robust deep-learning-based models for tasks. Here, we propose three geometric-learning-based models, GeoFitness, GeoDDG GeoDTm, score, ΔΔ<italic>G</italic> ΔT<sub>m</sub> mutations, respectively. In optimization designed novel...
Sanghamitra Dutta, Liang Ma, Tanay Kumar Saha, Di Liu, Joel Tetreault, Alejandro Jaimes. Proceedings of the Fifteenth Workshop on Graph-Based Methods for Natural Language Processing (TextGraphs-15). 2021.
In order to efficiently explore the chemical space of all possible small molecules, a common approach is compress dimension system facilitate downstream machine learning tasks. Towards this end, we present data driven for clustering potential energy landscapes molecular structures by applying recently developed Network Embedding techniques, obtain latent variables defined through embedding function. To scale up method, also incorporate an entropy sensitive adaptive scheme hierarchical...
Given the importance of ancient Chinese in capturing essence rich historical and cultural heritage, rapid advancements Large Language Models (LLMs) necessitate benchmarks that can effectively evaluate their understanding contexts. To meet this need, we present AC-EVAL, an innovative benchmark designed to assess advanced knowledge reasoning capabilities LLMs within context Chinese. AC-EVAL is structured across three levels difficulty reflecting different facets language comprehension: general...
<title>Abstract</title> Significant research progress has been made in the field of protein structure and fitness prediction. Particularly, single-sequence-based prediction methods like ESMFold OmegaFold achieve a balance between inference speed accuracy, showing promise for many downstream tasks. Here, we propose SPIRED, novel model that exhibits comparable performance to state-of-the-art but with approximately 5-fold acceleration at least one order magnitude reduction training consumption....
In order to efficiently explore the chemical space of all possible small molecules, a common approach is compress dimension system facilitate downstream machine learning tasks. Towards this end, we present data-driven for clustering potential energy landscapes molecular structures by applying recently developed Network Embedding techniques obtain latent variables defined through embedding function. To scale up method, also incorporate an entropy sensitive adaptive scheme hierarchical...
The pervasive application of Small Private Online Course (SPOC) provides a powerful impetus for the reform higher education. During teaching process, teacher needs to understand difficulty SPOC videos students in real time be more focused on difficulties and key points course flipped classroom. However, existing educational data mining techniques pay little attention video clustering or classification. In this paper, we propose an approach cluster based using video-watching SPOC....
Model scaling is an effective way to improve the accuracy of deep neural networks (DNNs) by increasing model capacity. However, existing approaches seldom consider underlying hardware, causing inefficient utilization hardware resources and consequently high inference latency. In this paper, we propose HACScale, a hardware-aware strategy fully exploit for higher accuracy. different dimensions DNNs are jointly scaled with consideration their contributions To efficiency width scaling, introduce...
Many chemical and biochemical systems can be intuitively modeled using networks. Due to the size complexity of many networks, we require tools for efficient network analysis. Of particular interest are techniques that embed vertices into vector spaces while preserving important properties original graph. In this article, {introduce a new method generating low-dimensional node embeddings directed graphs, random walk sampling methods feature learning on Additionally, demonstrate usefulness...
The signed network depicts individual cooperative or hostile attitude in a system. It is very important to study the characteristics of complex networks and predict attitudes by analyzing individuals their neighbors, which can divide into different modules communities. To detect networks, first, modularity function for utilized on basis existing function. Then, new module detection algorithm has also been put forward, high efficiency. Finally, applied both artificial real networks. results...
Brain-Machine Interfaces (BMI) establish a new information communication and control channel between the brain peripheral devices, which independent of spinal cord nervous system. In recent years, it has become hotspot research in field international intelligent science. However, current BMI usually belong to synchronous system, requires subjects perform specific tasks at program setting, greatly limits development The asynchronous attracted extensive attention researchers, whose study focus...