A5100742023- Topic Modeling
- Natural Language Processing Techniques
- Numerical methods for differential equations
- Electromagnetic Simulation and Numerical Methods
- Meteorological Phenomena and Simulations
- Multimodal Machine Learning Applications
- Advanced Graph Neural Networks
- Luminescence and Fluorescent Materials
- Matrix Theory and Algorithms
- Hydrological Forecasting Using AI
- Speech and dialogue systems
- Nanoplatforms for cancer theranostics
- Domain Adaptation and Few-Shot Learning
- Advanced Text Analysis Techniques
- Advanced Numerical Methods in Computational Mathematics
- Differential Equations and Numerical Methods
- Advanced Mathematical Physics Problems
- Advanced X-ray and CT Imaging
- Model Reduction and Neural Networks
- Recommender Systems and Techniques
- Emotion and Mood Recognition
- Fire Detection and Safety Systems
- Text and Document Classification Technologies
- Antimicrobial Peptides and Activities
- Climate variability and models
Qingdao University of Science and Technology
2025
Dalian University of Technology
2023-2024
Peking University
2017-2024
Inner Mongolia University
2023
Henan University of Technology
2022
Beijing University of Posts and Telecommunications
2021
Nanjing Normal University
2016-2018
Academy of Mathematics and Systems Science
2016
Bacterial infections have long been a formidable challenge in global public health, further compounded by the emergence of drug-resistant bacteria resulting from overuse and misuse antibiotics. Intelligent antibacterial strategies are garnering escalating attention concern due to their ability accurately recognize bacterial infections, efficiently eliminate pathogens, timely monitor infection end points order mitigate adverse effects excessive treatment on normal tissues. Hence, this study,...
In this paper, based on the theory of rooted trees and B-series, we propose concrete formulas substitution law for order = 5.With help new law, derive a B-series integrator extending averaged vector field (AVF) method to high order.The turns out be six exactly preserves energy Hamiltonian systems.Numerical experiments are presented demonstrate accuracy energy-preserving property sixth AVF method.
This study explores how to enhance the reasoning capabilities of large language models (LLMs) in knowledge base question answering (KBQA) by leveraging Monte Carlo Tree Search (MCTS). Semantic parsing-based KBQA methods are particularly challenging as these approaches require locating elements from bases and generating logical forms, demanding not only extensive annotated data but also strong capabilities. Although recent LLMs agents have demonstrated considerable potential, studies...
News modeling and user are the two core tasks of news recommendation. Accurate representation can enable recommendation system to provide users with precise services. Most existing methods use deep learning models such as CNN Self-Attention extract text features from titles abstracts generate specific vectors. However, CNN-based have fixed parameters cannot for different input words, while Self-Attention-based high computational costs difficult capture local effectively. In our proposed...
In this paper, we propose a deep spatio-temporal forecasting model (Deep-STF) for multi-site weather prediction post-processing by using both temporal and spatial information.In our proposed framework, the information is modeled CNN (convolutional neural network) module an encoder-decoder structure with attention mechanism.The novelty of work lies in that takes full account characteristics obtain forecasts multiple meteorological stations simultaneously same framework.We apply DeepSTF to...
Low-dose X-ray computed tomography (XCT) is a popular imaging technique to visualize the inside structure of object non-destructively. Model-based Iterative Reconstruction (MBIR) method can reconstruct high-quality image but at cost large computational demands. Therefore, MBIR ten resorts platforms with hardware accelerators such as GPUs speed up reconstruction process.
Large Language Models (LLMs) have shown prominent performance in various downstream tasks which prompt engineering plays a pivotal role optimizing LLMs' performance. This paper, not as an overview of current methods, aims to highlight the limitation designing prompts while holding anthropomorphic assumption that expects LLMs think like humans. From our review 35 representative studies, we demonstrate goal-oriented formulation, guides follow established human logical thinking, significantly...
Image-text retrieval, a fundamental cross-modal task, performs similarity reasoning for images and texts. The primary challenge image-text retrieval is semantic heterogeneity, where the features of visual textual modalities are rich but distinct. Scene graph an effective representation texts as it explicitly models objects their relations. Existing scene based methods have not fully taken regarding various granularities implicit in into consideration (e.g., triplets), inadequate feature...
In the subject recognition (SR) task under Knowledge Base Question Answering (KBQA), a common method is by training and employing general flat Named-Entity Recognition (NER) model. However, it not effective robust enough in case that recognized entity could be strictly matched to any subjects (KB). Compared NER models, nested models show more flexibility robustness tasks, whereas difficult employ model directly an SR task. this paper, we take advantage of features propose Improved Nested...
In practical applications, the raw input to a Knowledge Based Question Answering (KBQA) system may vary in forms, expressions, sources, etc. As result, actual contain various errors caused by noise data and processes of transmission, transformation, translation, it is significant evaluate enhance robustness KBQA model noisy questions. this paper, we generate 29 datasets questions based on original SimpleQuestions dataset model, propose which more robust Compared with traditional methods,...
This paper explores an efficient energy-preserving scheme for the coupled nonlinear Schrödinger system with emphasis on preserving Hamiltonian structure based weak formulation of system. First, is discretized in spatial direction by Galerkin spectral element method, and resulting semi-discrete rewritten as a finite-dimensional canonical Second, we apply energy method to discretize ordinary differential equations time obtain conservative scheme. Using error estimate without any restriction...
In code search, the Generation-Augmented Retrieval (GAR) framework, which generates exemplar snippets to augment queries, has emerged as a promising strategy address principal challenge of modality misalignment between and natural language particularly with demonstrated generation capabilities Large Language Models (LLMs). Nevertheless, our preliminary investigations indicate that improvements conferred by such an LLM-augmented framework are somewhat constrained. This limitation could...
Southeast Asia (SEA) is a region rich in linguistic diversity and cultural variety, with over 1,300 indigenous languages population of 671 million people. However, prevailing AI models suffer from significant lack representation texts, images, audio datasets SEA, compromising the quality for SEA languages. Evaluating challenging due to scarcity high-quality datasets, compounded by dominance English training data, raising concerns about potential misrepresentation. To address these...