A5113478345- Natural Language Processing Techniques
- Topic Modeling
- Semantic Web and Ontologies
- Genetic Mapping and Diversity in Plants and Animals
- Image and Signal Denoising Methods
- Dental Radiography and Imaging
- Medical Imaging Techniques and Applications
- Computational Drug Discovery Methods
- Smoking Behavior and Cessation
- Genetics, Bioinformatics, and Biomedical Research
- Various Chemistry Research Topics
- Genetic Associations and Epidemiology
- Advanced Measurement and Metrology Techniques
- Chemical Reactions and Isotopes
- Spectroscopy and Chemometric Analyses
- Advanced Vision and Imaging
- Advanced SAR Imaging Techniques
- Advanced X-ray and CT Imaging
- Advanced Image Processing Techniques
- Optical measurement and interference techniques
National University of Defense Technology
2024
University of Virginia
2007
Drug discovery is crucial for identifying candidate drugs various diseases.However, its low success rate often results in a scarcity of annotations, posing few-shot learning problem. Existing methods primarily focus on single-scale features, overlooking the hierarchical molecular structures that determine different properties. To address these issues, we introduce Universal Matching Networks (UniMatch), dual matching framework integrates explicit with implicit task-level via meta-learning,...
Decoding from the output distributions of large language models to produce high-quality text is a complex challenge in modeling. Various approaches, such as beam search, sampling with temperature, $k-$sampling, nucleus $p-$sampling, typical decoding, contrastive and have been proposed address this problem, aiming improve coherence, diversity, well resemblance human-generated text. In study, we introduce adaptive novel decoding strategy extending search by incorporating an degeneration...
Decoding strategies for large language models (LLMs) are a critical but often underexplored aspect of text generation tasks. Since LLMs produce probability distributions over the entire vocabulary, various decoding methods have been developed to transform these probabilities into coherent and fluent text, each with its own set hyperparameters. In this study, we present large-scale, comprehensive analysis how hyperparameter selection affects quality in open-ended across multiple LLMs,...
Open-ended text generation has become a prominent task in natural language processing due to the rise of powerful (large) models. However, evaluating quality these models and employed decoding strategies remains challenging because trade-offs among widely used metrics such as coherence, diversity, perplexity. Decoding methods often excel some while underperforming others, complicating establishment clear ranking. In this paper, we present novel ranking within multicriteria framework....
In the drug discovery process, low success rate of candidate screening often leads to insufficient labeled data, causing few-shot learning problem in molecular property prediction. Existing methods for prediction overlook sample selection bias, which arises from non-random chemical experiments. This bias data representativeness suboptimal performance. To overcome this challenge, we present a novel method named contextual representation anchor Network (CRA), where an refers cluster center...