模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2023 Vol.36 Issue.12, Published 2023-12-25
Adapative Perception and Learning of Open-Environment   
   
Adapative Perception and Learning of Open-Environment
0
LIU Chenglin
2023 Vol. 36 (12): 0-0 [Abstract] ( 207 ) [HTML 1KB] [ PDF 233KB] ( 951 )
1059 Research Advances on Theory of Open-Environment Machine Learning
YUAN Xiaotong, ZHANG Xuyao, LIU Xi, CHENG Zhen, LIU Chenglin
In an open environment, machine learning is faced with various challenges, including varying category sets, non-identically distributed data and noise interference. These challenges can result in a significant decline in the performance of traditional machine learning systems built under the closed-world assumption. Therefore, open-environment machine learning is a research focus on artificial intelligence. In this paper, the current status and recent important advances in the theoretical study of open-environment machine learning are discussed from the perspectives of generalization, optimization, robustness and performance measurement. For generalization theory, the advances on the generalization performance analysis of open-set detection, transfer/meta learning and sparse learning approaches are introduced. For optimization theory, the advances on the theoretical analysis of random and sparse optimization, online and continual optimization, as well as distributed and federated optimization approaches are introduced. For robustness theory, the advances on robust learning under adversarial samples, random noise and noisy labels are introduced. For performance measurement, a number of widely used performance measurement criterions for open-environment machine learning are introduced. Finally, some prospects on the theoretical research trends of open-environment machine learning are provided.
2023 Vol. 36 (12): 1059-1071 [Abstract] ( 337 ) [HTML 1KB] [ PDF 968KB] ( 1019 )
1072 Research Advances on Adaptive Perception and Learning in Changing Environment
ZHANG Xuyao, YUAN Xiaotong, LIU Chenglin
The research on artificial intelligence is gradually extended to open environment from closed environment. There are various changing factors in open environment leading to evident performance degradation of the traditional models and learning algorithms based on closed set assumption and independently and identically distributed assumption. Therefore, adaptive perception and learning in changing environments is a frontier topic in the field of artificial intelligence. The latest advances are introduced from three aspects. For category changing, research issues of open set recognition and out-of-distribution detection, new categories discovery and class-incremental learning are introduced. For data distribution changing, issues of domain adaptation, domain generalization and test-time adaptation are introduced. For data quality changing, issues of weakly supervised learning and label noise learning are introduced. Finally, future research trends are analyzed and discussed.
2023 Vol. 36 (12): 1072-1086 [Abstract] ( 209 ) [HTML 1KB] [ PDF 822KB] ( 978 )
1087 A Review on Architecture Adaptation of Neural Networks
LI Shu, QIN Xianping, ZHAI Xiaotong, ZHANG Long, ZHONG Guoqiang, XIANG Shiming
Network architecture adaptation aims to automatically design and optimize the neural network architectures based on specific learning tasks and data to meet the comprehensive needs of intelligent perception learning tasks in open environment. This paper is intended to provide a comprehensive review of network architecture adaptation methods. Firstly, the main methods of neural architecture search are elucidated and analyzed. Then, the research progress of network architecture adaptation is presented from three aspects: lightweight neural architecture search, intelligent perception tasks and continuous learning. On this basis, an adaptive evaluation index system of deep neural network components and architectures for open environment applications is established, and a network architecture adaptive method is proposed. Through the attention-guided micro-architecture adaptive mechanism and progressive discretization strategy, adaptive adjustment, optimization and gradual discretization of network structures are realized in the optimization process. The proposed method is compared with the existing methods. Finally, problems and challenges of current methods are discussed, and the future research directions are prospected.
2023 Vol. 36 (12): 1087-1103 [Abstract] ( 330 ) [HTML 1KB] [ PDF 1033KB] ( 986 )
1104 Progress in Attribution-Guided Adaptive Visual Perception and Structure Understanding
ZHANG Zhicheng, YANG Jufeng, CHENG Mingming, LIN Weiyao, TANG Jin, LI Chenglong, LIU Chenglin

Machines extract human-understandable information from the environment via adaptive perception to build intelligent system in open-world scenarios. Derived from the class-agnostic characteristics of attribute knowledge, attribution-guided perception methods and models are established and widely studied. In this paper, the tasks involved in attribution-guided adaptive visual perception and structure understanding are firstly introduced, and their applicable scenarios are analyzed. The representative research on four key aspects is summarized. Basic visual attribute knowledge extraction methods cover low-level geometric attributes and high-level cognitive attributes. Attribute knowledge-guided weakly-supervised visual perception includes weakly supervised learning and unsupervised learning under data label restrictions. Image self-supervised learning covers self-supervise contrastive learning and unsupervised commonality learning. Structured representation and understanding of scene images and their applications are introduced as well. Finally, challenges and potential research directions are discussed, such as the construction of large-scale benchmark datasets with multiple attributes, multi-modal attribute knowledge extraction, scene generalization of attribute knowledge perception models, the development of lightweight attribute knowledge-guided models and the practical applications of scene image representation.
2023 Vol. 36 (12): 1104-1126 [Abstract] ( 241 ) [HTML 1KB] [ PDF 2207KB] ( 565 )
1127 A Survey on Knowledge-Driven Multimodal Semantic Understanding
ZHENG Yihao, GUO Yijun, WU Lifang, HUANG Yan
Multimodal learning methods based on deep learning model achieve excellent semantic understanding performance in static, controllable and simple scenarios. However, their generalization ability in dynamic, open and other complex scenarios is still unsatisfactory. Human-like knowledge is introduced into multimodal semantic understanding methods in recent research, yielding impressive results. To gain deeper understanding of the current research progress in knowledge-driven multimodal semantic understanding, two main types of multimodal knowledge representation frameworks are summarized based on systematic investigation and analysis of relevant methods in this paper. The two main types of multimodal knowledge representation frameworks are relational and aligned, respectively. Several representative applications are discussed, including image-text matching, object detection, semantic segmentation, and vision-and-language navigation. In addition, the advantages and disadvan-tages of the current methods and the possible development trend in the future are concluded.
2023 Vol. 36 (12): 1127-1138 [Abstract] ( 305 ) [HTML 1KB] [ PDF 1101KB] ( 1137 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech 　Email:support@magtech.com.cn