A5100731925- Natural Language Processing Techniques
- Fault Detection and Control Systems
- Generative Adversarial Networks and Image Synthesis
- Advanced Computational Techniques and Applications
- Anomaly Detection Techniques and Applications
- Topic Modeling
- Text and Document Classification Technologies
- AI in cancer detection
- Time Series Analysis and Forecasting
- Mineral Processing and Grinding
- Explainable Artificial Intelligence (XAI)
- Cell Image Analysis Techniques
- Speech and dialogue systems
- Educational Technology and Pedagogy
- Data Stream Mining Techniques
- Adversarial Robustness in Machine Learning
- Handwritten Text Recognition Techniques
- Speech Recognition and Synthesis
- Machine Learning and Data Classification
- Machine Learning in Materials Science
- Music and Audio Processing
- AI and Multimedia in Education
- Neural Networks and Reservoir Computing
- Imbalanced Data Classification Techniques
- Higher Education and Teaching Methods
IBM (United States)
2021-2024
Minzu University of China
2013-2024
Beijing University of Chemical Technology
2022-2024
Jiangxi Normal University
2022-2023
Ministry of Education of the People's Republic of China
2022-2023
Lanzhou University of Technology
2023
University of California, Irvine
2021
Heilongjiang Vocational College of Art
2020
University of Central Florida
2020
University of Shanghai for Science and Technology
2019
Backpropagation-based visualizations have been proposed to interpret convolutional neural networks (CNNs), however a theory is missing justify their behaviors: Guided backpropagation (GBP) and deconvolutional network (DeconvNet) generate more human-interpretable but less class-sensitive than saliency map. Motivated by this, we develop theoretical explanation revealing that GBP DeconvNet are essentially doing (partial) image recovery which unrelated the decisions. Specifically, our analysis...
The computer vision world has been re-gaining enthusiasm in various pre-trained models, including both classical ImageNet supervised pre-training and recently emerged self-supervised such as simCLR [10] MoCo [40]. Pre-trained weights often boost a wide range of downstream tasks classification, detection, segmentation. Latest studies suggest that benefits from gigantic model capacity [11]. We are hereby curious ask: after pre-training, does indeed have to stay large for its transferability?...
Fault diagnosis plays an important role in ensuring process safety. It is noted that imbalance between fault data and normal always exists, multifault obviously outranges a single common, which leads to more challenges diagnosis. In this article, imbalanced method based on bias weights AdaBoost (BW-AdaBoost) proposed. First, majority samples are under-sampled by K-nearest neighbor (KNN) collect the boundary minority samples, then datasets formed samples. Second, different weak classifiers...
Novelty detection, or one-class classification, is of particular use in the analysis high-integrity systems, which examples failure are rare comparison with number stable behaviour, such that a conventional multi-class classification approach cannot be taken. Support Vector Machines (SVMs) popular means performing novelty and it practice to train-validate-test approach, often involving cross-validation, train SVM, then select appropriate values for its parameters. An alternative method, used...
3-Hydroxy-3-methylglutaryl coenzyme A reductase (HMGR) is a primary target in the current clinical treatment of hypercholesterolemia with specific inhibitors "statin" family. Statins are excellent class I (human) enzyme but relatively poor II enzyme, which well-known as potential to discover drugs fighting against invasive diseases originated from S. pneumoniae . However, no significantly effective HMGR have been reported so far. In present study, reasonable three-dimensional (3D) structure...
Modern automatic speech recognition (ASR) model is required to accurately transcribe diverse signals (from different domains, languages, accents, etc) given the specific contextual information in various application scenarios. Classic end-to-end models fused with extra language perform well, but mainly data matching scenarios and are gradually approaching a bottleneck. In this work, we introduce Seed-ASR, large (LLM) based model. Seed-ASR developed on framework of audio conditioned LLM...
In order to segment nucleus and cytoplasm of single cervical cell in the semear images accurately, a novel Gradient Vector Field (GVF) Snake is proposed by improving edge map GVF (PGVF) enhance capacity active contour model gradient vector field locate boundary nuclei cytoplasmic. our method, image converted into polar coordinates Sobel operator used calculate horizontal boundary, then Sand Inhibition Method designed inhibite influence interference elements on location, finally cells are...
Abstract In this era of rapid development, the level human science and technology is constantly improving. With continuous development times progress society, big data has been widely used in our daily production life. The research College English online learning platform model a hot topic teaching many college majors. order to better study platform, colleges universities put forward method using analyze network model, through Based on prediction e-learning combined with new requirements for...
We show an engineered dissipation for the spin squeezing and light entanglement in a quantum beat system, which two bright fields interact with ensemble of three-level atoms V configuration. The is based on atom-field nonlinear interaction that controlled by atomic coherence between excited states off two-photon resonance. Physical analysis numerical verification are presented symmetrical parameters using dressed states. It shown particular parameters, induces almost perfect two-mode both...
Automatic error detection and correction of text is an important research area natural language processing. Studies abroad originated in 1960s on automatic correction. Since the 1990s, many academic studies have been made China. This paper analyzes state art check correct approaches for English Chinese text.
In this paper, we propose an improved double-stage classfier for online recognition of handwritten chemical symbols. the first stage, SVM based classifier is used to roughly classify symbols into Non-Ring Structure(NRS) and Organic Ring Structure(ORS). Then, HMM fine classification at second stage. During use frequency domain feature instead commonly Geometrical or Statistical perform task. addition, improve accuracy ORS speed processing, a MPSR algorithm. Finally, achieve top-1 88.95% top-3...
Abstract In actual industrial processes, although a large number of original data are easy to obtain, only few samples effectively labelled, which is insufficient construct supervised fault diagnostic model. Facing the demand diagnosis, in this paper, novel density ratio (DR)‐based batch active learning (BAL) diagnosis method integrated with adaptive Laplacian graph trimming (ALGT) proposed. First, under framework, new index DR‐based on local reachability (LRD) proposed search low and high...
Table question answering is a popular task that assesses model's ability to understand and interact with structured data. However, the given table often does not contain sufficient information for question, necessitating integration of external knowledge. Existing methods either convert both knowledge into text, which neglects nature table; or they embed queries sources in interaction table, complicates process. In this paper, we propose simple yet effective method integrate table. Our first...
<title>Abstract</title> The preservation and restoration of the Dunhuang murals, a paramount element China's ancient artistic heritage, are confronted with significant challenges due to their extensive degradation over centuries. Traditional inpainting methods, which mainly focused on spatial domain, frequently fall short in authentically restoring these especially considering distinct style. In response challenges, our study introduces novel wavelet-based dual-stream network (WDSN). This...