A5101647485- Medical Image Segmentation Techniques
- Face and Expression Recognition
- Reinforcement Learning in Robotics
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Handwritten Text Recognition Techniques
- Remote-Sensing Image Classification
- Face recognition and analysis
- Advanced Neuroimaging Techniques and Applications
- Text and Document Classification Technologies
- Blind Source Separation Techniques
- Sparse and Compressive Sensing Techniques
- Game Theory and Applications
- Advanced Vision and Imaging
- Advanced MRI Techniques and Applications
- AI in cancer detection
- Domain Adaptation and Few-Shot Learning
- Vehicle License Plate Recognition
- Topic Modeling
- Neural Networks and Applications
- Natural Language Processing Techniques
- Artificial Intelligence in Games
- Brain Tumor Detection and Classification
- Biometric Identification and Security
East China Normal University
2015-2024
Shanghai Jiao Tong University
2006-2024
AviChina Industry & Technology (China)
2022
University College London
2016-2020
Shanghai Children's Hospital
2019
Peking University
2016
Southern Medical University
2016
UCL Australia
2016
China Post (China)
2015
Columbia University
2010-2012
Predicting user responses, such as clicks and conversions, is of great importance has found its usage inmany Web applications including recommender systems, websearch online advertising. The data in those applicationsis mostly categorical contains multiple fields, a typicalrepresentation to transform it into high-dimensional sparsebinary feature representation via one-hot encoding. Facing withthe extreme sparsity, traditional models may limit their capacityof mining shallow patterns from the...
An algorithm for license plate recognition (LPR) applied to the intelligent transportation system is proposed on basis of a novel shadow removal technique and character algorithms. This paper has two major contributions. One contribution new binary method, i.e., which based improved Bernsen combined with Gaussian filter. Our second known as support vector machine (SVM) integration. In SVM integration, features are extracted from elastic mesh, entire address string taken object study, opposed...
Convolutional Neural Networks (CNNs) have achieved remarkable performance in various computer vision tasks but this comes at the cost of tremendous computational resources, partly due to convolutional layers extracting redundant features. Recent works either compress well-trained large-scale models or explore well-designed lightweight models. In paper, we make an attempt exploit spatial and channel redundancy among features for CNN compression propose efficient convolution module, called...
Modeling and learning of brain activity patterns represent a huge challenge to the brain-computer interface (BCI) based on electroencephalography (EEG). Many existing methods estimate uncorrelated instantaneous demixing EEG signals classify multiclass motor imagery (MI). However, condition uncorrelation does not hold true in practice, because regions work with partial or complete collaboration. This proposes novel method, termed as common Bayesian network (CBN), discriminate MI signals....
Multi-agent interaction is a fundamental aspect of autonomous driving in the real world. Despite more than decade research and development, problem how to competently interact with diverse road users scenarios remains largely unsolved. Learning methods have much offer towards solving this problem. But they require realistic multi-agent simulator that generates competent interactions. To meet need, we develop dedicated simulation platform called SMARTS (Scalable Multi-Agent RL Training...
Skin biopsy histopathological analysis is one of the primary methods used for pathologists to assess presence and deterioration melanoma in clinical. A comprehensive reliable pathological result correctly segmented its interaction with benign tissues, therefore providing accurate therapy. In this study, we applied deep convolution network on hyperspectral pathology images perform segmentation melanoma. To make best use spectral properties three dimensional data, proposed a 3D fully...
Predicting user responses, such as clicks and conversions, is of great importance has found its usage in many Web applications including recommender systems, web search online advertising. The data those mostly categorical contains multiple fields; a typical representation to transform it into high-dimensional sparse binary feature via one-hot encoding. Facing with the extreme sparsity, traditional models may limit their capacity mining shallow patterns from data, i.e. low-order...
Deep Q-learning has achieved significant success in single-agent decision making tasks. However, it is challenging to extend large-scale multi-agent scenarios, due the explosion of action space resulting from complex dynamics between environment and agents. In this paper, we propose make computation tractable by treating Q-function (w.r.t. state joint-action) as a high-order high-dimensional tensor then approximate with factorized pairwise interactions. Furthermore, utilize composite deep...
Abstract Offline reinforcement learning leverages previously collected offline datasets to learn optimal policies with no necessity access the real environment. Such a paradigm is also desirable for multi-agent (MARL) tasks, given combinatorially increased interactions among agents and However, in MARL, of pre-training online fine-tuning has not been studied, nor even or benchmarks MARL research are available. In this paper, we facilitate by providing large-scale using them examine usage...
There is a growing worldwide trend to implement livestock traceability systems. This paper aims explore how iris analysis and recognition can be utilised on cow identification enhance management in its system. In general, typical system based includes imaging, detection, recognition. First, the image quality of captured sequences assessed clear selected for subsequent process. Second, inner outer boundaries are fitted respectively as two ellipses edge images during segmentation. Then we get...
Recently, the rapid development of word embedding and neural networks has brought new inspiration to various NLP IR tasks. In this paper, we describe a staged hybrid model combining Recurrent Convolutional Neural Networks (RCNN) with highway layers. The network module is incorporated in middle takes output bi-directional Network (Bi-RNN) first stage provides (CNN) last input. experiment shows that our outperforms common models (CNN, RNN, Bi-RNN) on sentiment analysis task. Besides, how...
Humans are capable of attributing latent mental contents such as beliefs or intentions to others. The social skill is critical in daily life for reasoning about the potential consequences others' behaviors so plan ahead. It known that humans use ability recursively by considering what others believe their own beliefs. In this paper, we start from level-$1$ recursion and introduce a probabilistic recursive (PR2) framework multi-agent reinforcement learning. Our hypothesis it beneficial each...
Object detection on very high resolution (VHR) remote sensing images is a crucial task that has seen remarkable progress in developing deep learning techniques. However, learning-based methods rely heavily the quality and quantity of labeled data. Although few-shot object (FSOD) can mitigate this dependency, existing still face challenges, including domain shifts between base novel classes, misclassification due to class similarities, limited ability acquire effective information from few...
The evolution of large language models (LLMs) toward artificial superhuman intelligence (ASI) hinges on data reproduction, a cyclical process in which generate, curate and retrain novel to refine capabilities. Current methods, however, risk getting stuck reproduction trap: optimizing outputs within fixed human-generated distributions closed loop leads stagnation, as merely recombine existing knowledge rather than explore new frontiers. In this paper, we propose games pathway expanded...
While large language models (LLMs) have significantly advanced mathematical reasoning, Process Reward Models (PRMs) been developed to evaluate the logical validity of reasoning steps. However, PRMs still struggle with out-of-distribution (OOD) challenges. This paper identifies key OOD issues, including step OOD, caused by differences in patterns across model types and sizes, question which arises from dataset shifts between training data real-world problems. To address these we introduce...
Evaluating deep reinforcement learning (DRL) agents against targeted behavior attacks is critical for assessing their robustness. These aim to manipulate the victim into specific behaviors that align with attacker’s objectives, often bypassing traditional reward-based defenses. Prior methods have primarily focused on reducing cumulative rewards; however, rewards are typically too generic capture complex safety requirements effectively. As a result, focusing solely reward reduction can lead...