Article
Engineering, Civil
Haohan Yang, Haochen Liu, Zhongxu Hu, Anh-Tu Nguyen, Thierry-Marie Guerra, Chen Lv
Summary: This paper proposes a vision Transformer-enabled weakly supervised contrastive learning framework for accurate recognition of driver distraction. Experimental results demonstrate that the proposed approach has more accurate and robust performance in recognizing unknown driver activities.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xudong Tang, Chao Dong, Wei Zhang
Summary: In the era of User Generated Content (UGC), the influence of authors on text clustering remains largely underexplored. To address this issue, we propose a novel Contrastive Author-aware Text clustering approach, CAT, which incorporates author information through contrastive learning and multi-view representations, resulting in improved clustering performance.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Wenhui Zhao, Quanxue Gao, Shikun Mei, Ming Yang
Summary: This paper introduces a method for self-representation subspace learning and proposes a Contrastive Self-representation model for Clustering (CSC) that considers the similarity/dissimilarity between positive/negative pairs and achieves sparsity in the coefficient matrix through regularizer, to better describe the cluster structure. Experimental results demonstrate the superiority of the proposed method.
Article
Computer Science, Artificial Intelligence
Jun Yin, Haowei Wu, Shiliang Sun
Summary: As an essential part of unsupervised learning, deep clustering is becoming increasingly important with the development of deep neural networks. The combination of contrastive learning and deep clustering has achieved more competitive performance. However, previous studies mostly used random augmentations for contrastive clustering, which may lead to false positives and ignore the semantic variations of different samples. To overcome these limitations, we propose a novel framework called Contrastive Clustering with Effective Sample pairs construction (CCES), which leverages an effective data augmentation method and constructs positive sample pairs based on nearest-neighbor mining. Experimental results on four challenging datasets demonstrate that CCES outperforms state-of-the-art methods.
INFORMATION FUSION
(2023)
Article
Computer Science, Information Systems
Xingyu Xie, Lei Zhang, Yan Wang, Zizhou Wang, Yu Hua
Summary: This paper explores the relationship between batch size and feature decorrelation in contrastive learning, and proposes a method to indirectly promote feature decorrelation by utilizing the linear independence created by contrastive learning with small batch sizes. It also introduces a bi-vector-based contrastive learning method, which is of interest to researchers with limited resources. Experimental results demonstrate the efficacy of the proposed method against state-of-the-art methods on five benchmark datasets.
INFORMATION SCIENCES
(2023)
Article
Automation & Control Systems
Chao Huang, Zhihao Wu, Jie Wen, Yong Xu, Qiuping Jiang, Yaowei Wang
Summary: In this article, a novel method called TAC-Net is proposed to address the problem of anomaly detection in intelligent video surveillance. It calculates anomaly scores by utilizing contrastive similarity and employs deep contrastive self-supervised learning in multiple self-supervised tasks to capture high-level semantic features.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Wei Xia, Tianxiu Wang, Quanxue Gao, Ming Yang, Xinbo Gao
Summary: This article addresses challenges in multi-modal clustering methods based on deep neural networks. It proposes a novel Graph Embedding Contrastive Multi-modal Clustering network (GECMC) that integrates representation learning and multi-modal clustering. GECMC effectively maximizes intra-cluster similarities and minimizes inter-cluster similarities, and can handle out-of-sample data. Experimental results demonstrate that GECMC outperforms 14 competitive methods on four challenging datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Yecheng Guo, Liang Bai, Xian Yang, Jiye Liang
Summary: This article introduces a new unsupervised clustering model, which improves clustering results for images by integrating pairwise constraints into the clustering process. The model autonomously learns pairwise constraints, eliminating the need for labeled images and offering a practical solution to the challenge of insufficient supervised information in unsupervised clustering tasks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Transportation Science & Technology
Kehua Chen, Jindong Han, Siyuan Feng, Meixin Zhu, Hai Yang
Summary: This article studies the issue of driver profiling in ride-hailing services and proposes a Hierarchical Graph Contrastive Learning (HGCL) framework that automatically learns low-dimensional embeddings from raw GPS data to encode driver behaviors. Experimental results demonstrate the efficacy of the proposed framework in driver profiling.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2023)
Article
Computer Science, Artificial Intelligence
Chaoyang Xu, Renjie Lin, Jinyu Cai, Shiping Wang
Summary: This paper proposes a new deep image clustering framework that combines contrastive learning with neighbor relation mining, achieving more semantic meaningful representations and accurate image clusters. The framework alternates between contrastive learning and neighbor relation mining to update the model.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Tong Wang, Junhua Wu, Yaolei Qi, Xiaoming Qi, Juwei Guan, Yuan Zhang, Guanyu Yang
Summary: This study proposes a neighborhood contrastive representation learning method, NCAGC, for attributed graph clustering, which improves clustering performance through representation learning of similar nodes and learning of a self-expression coefficient matrix.
Article
Computer Science, Artificial Intelligence
Zihua Zhao, Rong Wang, Zheng Wang, Feiping Nie, Xuelong Li
Summary: Graph clustering based on graph contrastive learning (GCL) is a dominant paradigm in the current research field. However, GCL-based methods often produce false negative samples, impacting the learned representations and limiting clustering performance. To address this issue, we propose maintaining mutual information (MI) between representations and inputs to mitigate the loss of semantic information. We also introduce a graph clustering method that penalizes GCL with reconstruction error, improving clustering performance with a carefully designed reconstruction decoder.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xin Peng, Jieren Cheng, Xiangyan Tang, Jingxin Liu, Jiahua Wu
Summary: Graph representation is essential in graph clustering. Contrastive learning has become a popular and powerful paradigm, aiming to maximize the mutual information between augmented graph views. However, existing literature tends to collapse representation, resulting in less discriminative graph representations. To address this, we propose a novel self-supervised learning method called dual contrastive learning network (DCLN), which reduces redundant information in a dual manner. Our method outperforms state-of-the-art methods, as demonstrated by extensive experiments on six benchmark datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Information Systems
Yiming Wang, Dongxia Chang, Zhiqiang Fu, Jie Wen, Yao Zhao
Summary: This paper proposes an augmentation-free graph contrastive learning framework to solve the problem of partial multi-view clustering. By constructing relation graphs and transferring graphs, the proposed approach elevates instance-level contrastive learning and missing data inference to the cluster-level, effectively improving the clustering performance.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Ning Huyan, Dou Quan, Xiangrong Zhang, Xuefeng Liang, Jocelyn Chanussot, Licheng Jiao
Summary: This paper investigates the issue of outlier detection in deep learning methods and proposes an unsupervised outlier detection method using a memory module and a contrastive learning module. Extensive experiments demonstrate that the proposed method performs well on four benchmark datasets and outperforms eleven state-of-the-art methods.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)