检索结果-内蒙古大学图书馆

Is Dataset Quality Still a Concern in Diagnosis Using Large Foundation Model?

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Lin, Ziqin Li, Heng Li, Zinan Fu, Huazhu Liu, Jiang Research Institute of Trustworthy Autonomous Systems Southern University of Science and Technology Shenzhen China Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China Institute of High Performance Computing Agency for Science Technology and Research Singapore Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Southern University of Science and Technology Shenzhen China

Recent advancements in pre-trained large foundation models (LFM) have yielded significant breakthroughs across various domains, including natural language processing and computer vision. These models have been particularly impactful in the domain of medical diagnostic tasks. With abundant unlabeled data, an LFM has been developed for fundus images using the Vision Transformer (VIT) and a self-supervised learning framework. This LFM has shown promising performance in fundus disease diagnosis across multiple datasets. On the other hand, deep learning models have long been challenged by dataset quality issues, such as image quality and dataset bias. To investigate the influence of data quality on LFM, we conducted explorations in two fundus diagnosis tasks using datasets of varying quality. Specifically, we explored the following questions: Is LFM more robust to image quality? Is LFM affected by dataset bias? Can fine-tuning techniques alleviate these effects? Our investigation found that LFM exhibits greater resilience to dataset quality issues, including image quality and dataset bias, compared to typical convolutional networks. Furthermore, we discovered that overall fine-tuning is an effective adapter for LFM to mitigate the impact of dataset quality issues. Copyright © 2024, The Authors. All rights reserved.

关键词： Image quality

Multi-Objective Archiving

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Li, Miqing López-Ibáñez, Manuel Yao, Xin The School of Computer Science University of Birmingham United Kingdom Alliance Manchester Business School University of Manchester United Kingdom The Research Institute of Trustworthy Autonomous Systems Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China

Most multi-objective optimisation algorithms maintain an archive explicitly or implicitly during their search. Such an archive can be solely used to store high-quality solutions presented to the decision maker, but in many cases may participate in the search process (e.g., as the population in evolutionary computation). Over the last two decades, archiving, the process of comparing new solutions with previous ones and deciding how to update the archive/population, stands as an important issue in evolutionary multi-objective optimisation (EMO). This is evidenced by constant efforts from the community on developing various effective archiving methods, ranging from conventional Pareto-based methods to more recent indicator-based and decomposition-based ones. However, the focus of these efforts is on empirical performance comparison in terms of specific quality indicators;there is lack of systematic study of archiving methods from a general theoretical perspective. In this paper, we attempt to conduct a systematic overview of multiobjective archiving, in the hope of paving the way to understand archiving algorithms from a holistic perspective of theory and practice, and more importantly providing a guidance on how to design theoretically desirable and practically useful archiving algorithms. In doing so, we also present that archiving algorithms based on weakly Pareto compliant indicators (e.g., ϵ-indicator), as long as designed properly, can achieve the same theoretical desirables as archivers based on Pareto compliant indicators (e.g., hypervolume indicator). Such desirables include the property limit-optimal, the limit form of the possible optimal property that a bounded archiving algorithm can have with respect to the most general form of superiority between solution sets. Copyright © 2023, The Authors. All rights reserved.

关键词： Multiobjective optimization

Examination of the Multimodal Nature of Multi-Objective Neural Architecture Search

学校读者我要写书评

暂无评论

Examination of the Multimodal Nature of Multi-Objective Neur...

IEEE Symposium Series on Computational Intelligence (SSCI)

作者： Cheng Gong Yang Nan Lie Meng Pang Hisao Ishibuchi Qingfu Zhang Department of Computer Science and Engineering Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Southern University of Science and Technology Shenzhen China Department of Computer Science City University of Hong Kong Kowloon Tong Hong Kong The City University of Hong Kong Shenzhen Research Institute Shenzhen China

Remarkable successes in deep learning have spurred significant growth in the field of neural architecture search (NAS), which is rapidly advancing as a promising technique for automating the design of network architecture. From an optimization standpoint, a NAS task for a given search space can be viewed as a multi-objective optimization problem (MOP) when considering multiple design criteria simultaneously (e.g., prediction accuracy, architecture complexity, hardware efficiency). However, whether a NAS problem is a multimodal multi-objective optimization problem or not (i.e., whether a single non-dominated solution in the objective space has multiple different neural network architectures or not) has not been examined in the literature. This presents an intriguing research question that merits further investigation. To fill this gap, we examine the multimodal nature of seven multi-objective NAS problems. By doing so, this work aims to help MOP researchers to better understand the characteristics of the multi-objective NAS problems.

关键词：

Distortion-adaptive salient object detection in 360◦ omnidirectional images

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Li, Jia Su, Jinming Xia, Changqun Tian, Yonghong State Key Laboratory of Virtual Reality Technology and Systems School of Computer Science and Engineering Beihang University Beijing100191 China Beijing Adavanced Innovation Center for Big Data and Brain Computing Beihang University China National Engineering Laboratory for Video Technology School of Electronics Engineering and Computer Science Peking University Beijing China Peng Cheng Laboratory Shenzhen518000 China

Image-based salient object detection (SOD) has been extensively explored in the past decades. However, SOD on 360◦ omnidirectional images is less studied owing to the lack of datasets with pixel-level annotations. Toward this end, this paper proposes a 360◦ image-based SOD dataset that contains 500 high-resolution equirectangular images. We collect the representative equirectangular images from five mainstream 360◦ video datasets and manually annotate all objects and regions over these images with precise masks with a free-viewpoint way. To the best of our knowledge, it is the first public available dataset for salient object detection on 360◦ scenes. By observing this dataset, we find that distortion from projection, large-scale complex scene and small salient objects are the most prominent characteristics. Inspired by these foundings, this paper proposes a baseline model for SOD on equirectangular images. In the proposed approach, we construct a distortion-adaptive module to deal with the distortion caused by the equirectangular projection. In addition, a multiscale contextual integration block is introduced to perceive and distinguish the rich scenes and objects in omnidirectional scenes. The whole network is organized in a progressively manner with deep supervision. Experimental results show the proposed baseline approach outperforms the top-performanced state-of-the-art methods on 360◦ SOD dataset. Moreover, benchmarking results of the proposed baseline approach and other methods on 360◦ SOD dataset show the proposed dataset is very challenging, which also validate the usefulness of the proposed dataset and approach to boost the development of SOD on 360◦ omnidirectional scenes. Copyright © 2019, The Authors. All rights reserved.

关键词： Benchmarking

Precise nanomedicine for intelligent therapy of cancer

学校读者我要写书评

暂无评论

science China Chemistry 2018年第12期61卷 1503-1552页

作者： Huabing Chen Zhanjun Gu Hongwei An Chunying Chen Jie Chen Ran Cui Siqin Chen Weihai Chen Xuesi Chen Xiaoyuan Chen Zhuo Chen Baoquan Ding Qian Dong Qin Fan Ting Fu Dayong Hou Qiao Jiang Hengte Ke Xiqun Jiang Gang Liu Suping Li Tianyu Li Zhuang Liu Guangjun Nie Muhammad Ovais Daiwen Pang Nasha Qiu Youqing Shen Huayu Tian Chao Wang Hao Wang Ziqi Wang Huaping Xu Jiang-Fei Xu Xiangliang Yang Shuang Zhu Xianchuang Zheng Xianzheng Zhang Yanbing Zhao Weihong Tan Xi Zhang Yuliang Zhao State Key Laboratoiy of Radiation Medicine and Protection Jiangsu Key Laboratoiy of Neuropsychiatric Diseases College of Pharmaceutical Sciences Soochow University Suzhou 215123 China Key Laboratoiy of Biomedical Polymers of Ministry of Education Department of Chemistiy Wuhan University Wuhan 430072 China Laboratoiy of Molecular Imaging and Nanomedicine National Institute of Biomedical Imaging and Bioengineering National Institutes of Health Bethesda MD 20892 USA Molecular Science and Biomedicine Laboratoiy State Key Laboratory of Chemo/Bio-Sensing and Chemometrics College of Chemistiy and Chemical Engineering Aptamer Engineering Center of Hunan Province Hunan University Changsha 410082 China Jiangsu Key Laboratoiy for Carbon-Based Functional Materials & Devices Institute of Functional Nano & Soft Materials Soochow University Suzhou 215123 China Department of Polymer Science and Engineering College cf Chemistry and Chemical Engineering Nanjing University Nanjing 210093 China State Key Laboratory of Molecular Vaccinology and Molecular Diagnostics & Center for Molecular Imaging and Translational Medicine School of Public Health Xiamen University Xiamen 361102 China Department of Chemistiy Tsinghua University Beijing 100084 China National Engineering Research Center for Nanomedicine College of Life Science and Technology Huazhong University of Science and Technology Wuhan 430074 China Department of Chemistiy and Department of Physiology and Functional Genomics Center for Research at Bio/nano Interface Health Cancer Center UF Genetics Institute and McKnight Brain Institute University of Florida Gainesville FL 32611-7200 USA institute of High Energy Physics Chinese Academy of Sciences Beijing 100049 China CAS Key Laboratoiy for Biomedical Effects of Nanomaterials and Nanosafety CAS Center for Excellence in Nanoscience National Center for Nanoscience and Technology of China Beijing 100190 China University of Chinese Academy of Sciences Beijing 100049 China Key Labo

Precise nanomedicine has been extensively explored for efficient cancer imaging and targeted cancer therapy, as evidenced by a few breakthroughs in their preclinical and clinical explorations. Here, we demonstrate the recent advances of intelligent cancer nanomedicine, and discuss the comprehensive understanding of their structure-function relationship for smart and efficient cancer nanomedicine including various imaging and therapeutic applications, as well as nanotoxicity. In particular, a few emerging strategies that have advanced cancer nanomedicine are also highlighted as the emerging focus such as tumor imprisonment, supramolecular chemotherapy, and DNA nanorobot. The challenge and outlook of some scientific and engineering issues are also discussed in future development. We wish to highlight these new progress of precise nanomedicine with the ultimate goal to inspire more successful explorations of intelligent nanoparticles for future clinical translations.

关键词： nanomedicine nanoparticles cancer therapy cancer imaging intelligent therapy

Personalized Sleep Staging Leveraging Source-free Unsupervised Domain Adaptation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhou, Yangxuan Zhao, Sha Wang, Jiquan Jiang, Haiteng Li, Shijian Luo, Benyan Li, Tao Pan, Gang State Key Laboratory of Brain-machine Intelligence Zhejiang University Hangzhou China College of Computer Science and Technology Zhejiang University Hangzhou China Department of Neurobiology Affiliated Mental Health Center Hangzhou Seventh People’s Hospital Zhejiang University School of Medicine Hangzhou China MOE Frontier Science Center for Brain Science and Brain-machine Integration Zhejiang University Hangzhou China The First Affiliated Hospital College of Medicine Zhejiang University Hangzhou China

Sleep staging is important for monitoring sleep quality and diagnosing sleep-related disorders. Recently, numerous deep learning-based models have been proposed for automatic sleep staging using polysomnography recordings. Most of them are trained and tested on the same labeled datasets which results in poor generalization to unseen target domains. However, they regard the subjects in the target domains as a whole and overlook the individual discrepancies, which limits the model’s generalization ability to new patients (i.e., unseen subjects) and plug-and-play applicability in clinics. To address this, we propose a novel Source-Free Unsupervised Individual Domain Adaptation (SF-UIDA) framework for sleep staging, leveraging sequential cross-view contrasting and pseudo-label based fine-tuning. It is actually a two-step subject-specific adaptation scheme, which enables the source model to effectively adapt to newly appeared unlabeled individual without access to the source data. It meets the practical needs in real-world scenarios, where the personalized customization can be plug-and-play applied to new ones. Our framework is applied to three classic sleep staging models and evaluated on three public sleep datasets, achieving the state-of-the-art performance. Copyright © 2024, The Authors. All rights reserved.

关键词： Sleep research

Emergence and reconfiguration of modular structure for synaptic neural networks during continual familiarity detection

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Gu, Shi Mattar, Marcelo G. Tang, Huajin Pan, Gang School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China Shenzhen Institute for Advanced Study University of Electronic Science and Technology of China Shenzhen China Psychology Department New York University United States College of Computer Science and Technology Zhejiang University Hangzhou China State Key Laboratory of Brain Machine Intelligence Zhejiang University Hangzhou China

While advances in artificial intelligence and neuroscience have enabled the emergence of neural networks capable of learning a wide variety of tasks, our understanding of the temporal dynamics of these networks remains limited. Here, we study the temporal dynamics during learning of Hebbian Feedforward (HebbFF) neural networks in tasks of continual familiarity detection. Drawing inspiration from the field of network neuroscience, we examine the network's dynamic reconfiguration, focusing on how network modules evolve throughout learning. Through a comprehensive assessment involving metrics like network accuracy, modular flexibility, and distribution entropy across diverse learning modes, our approach reveals various previously unknown patterns of network reconfiguration. In particular, we find that the emergence of network modularity is a salient predictor of performance, and that modularization strengthens with increasing flexibility throughout learning. These insights not only elucidate the nuanced interplay of network modularity, accuracy, and learning dynamics but also bridge our understanding of learning in artificial and biological realms. © 2023, CC BY.

关键词： Feedforward neural networks

Keiki: Towards realistic danmaku generation via sequential GANs

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wang, Ziqi Liu, Jialin Yannakakis, Georgios N. Research Institute of Trustworthy Autonomous System Southern University of Science and Technology Shenzhen China Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen China Institute of Digital Games University of Malta Msida2080 Malta

Search-based procedural content generation methods have recently been introduced for the autonomous creation of bullet hell games. Search-based methods, however, can hardly model patterns of danmakus—the bullet hell shooting entity—explicitly and the resulting levels often look non-realistic. In this paper, we present a novel bullet hell game platform named Keiki, which allows the representation of danmakus as a parametric sequence which, in turn, can model the sequential behaviours of danmakus. We employ three types of generative adversarial networks (GANs) and test Keiki across three metrics designed to quantify the quality of the generated danmakus. The time-series GAN and periodic spatial GAN show different yet competitive performance in terms of the evaluation metrics adopted, their deviation from human-designed danmakus, and the diversity of generated danmakus. The preliminary experimental studies presented here showcase that potential of time-series GANs for sequential content generation in games. Copyright © 2021, The Authors. All rights reserved.

关键词： Generative adversarial networks

Initial Populations with a Few Heuristic Solutions Significantly Improve Evolutionary Multi-Objective Combinatorial Optimization

学校读者我要写书评

暂无评论

Initial Populations with a Few Heuristic Solutions Significa...

IEEE Symposium Series on Computational Intelligence (SSCI)

Population initialization is a crucial and essential step in evolutionary multi-objective optimization (EMO) algorithms. The quality of the generated initial population can significantly affect the performance of an EMO algorithm. However, few studies have focused on designing a generalized initialization method to improve the performance of EMO algorithms in solving multi-objective combinatorial optimization (MOCO) problems. Most of the existing advanced initialization methods involve complex techniques tailored to the specific characteristics of the problems to be solved. In this paper, we propose a general and effective framework of population initialization for EMO algorithms, aiming to improve their performances in solving various MOCO problems. Our approach involves the inclusion of a few specific heuristic solutions, including extreme solutions and a center solution, into the initial population. This inclusion serves to guide the evolution of the population throughout the optimization process. Our experimental results show that initial populations with a few heuristic solutions significantly improve the performance of EMO algorithms. Algorithm behavior analysis and further study are also provided, allowing for a comprehensive understanding of the effectiveness and applicability of our proposed method.

关键词：