📝 Publications

Xinhang Wan^*, Dongqiang Gou^*, Xinwang Liu, En Zhu, Xuming He. PDF

• We propose a novel approach that learns an affordance-aware 3D representation and employs a stage-wise inference strategy leveraging the dependency between grounding and classification tasks.

Xinhang Wan, Jiyuan Liu, Qian Qu, Suyuan Liu, Chuyu Zhang, Fangdi Wang, Xinwang Liu^#, En Zhu^#, Kunlun He. Code PDF

• To the best of our knowledge, this is the first attempt to address novel class discovery in the multi-view setting. We exploit intra-view and inter-view correlations to transfer knowledge from known classes to novel classes.

Xinhang Wan, Jiyuan Liu, Xinwang Liu^#, Yi Wen, Hao Yu, Siwei Wang, Shengju Yu, Tianjiao Wan, Jun Wang, En Zhu^#. Code PDF

• We propose an efficient algorithm to tackle the sample labeling task in semi-supervised multi-view learning. The samples of low classification confidence are labeled as high priorities.

Xinhang Wan, Bin Xiao, Xinwang Liu^#, Jiyuan Liu, Weixuan Liang, En Zhu^#. Code PDF

• We study a new paradigm for large-scale multi-view clustering called the incomplete continual data problem (ICDP) and propose FCMVC-IV to tackle the problem.

Xinhang Wan, Jiyuan Liu, Hao Yu, Qian Qu, Ao li, Xinwang Liu^#, Ke Liang, Zhibin Dong, En Zhu^#. Code PDF

• We study a new paradigm on continual multi-view clustering, termed catastrophic forgetting problem (CFP). A clustering then sample strategy is deployed to extract and update the filtered structure information of prior views, then the attained information will guide the clustering when a new view arrives.

Xinhang Wan, Jiyuan Liu, Xinbiao Gan, Xinwang Liu^#, Siwei Wang, Yi Wen, Tianjiao Wan, En Zhu^#. Code PDF

• By directly calculating the distance among samples with diverse representation, we incorporate matrix factorization and k-means into a unified framework with linear complexity. They negotiate with each other and boost the clustering performance.

Xinhang Wan, Xinwang Liu^#, Jiyuan Liu, Siwei Wang, Yi Wen, Weixuan Liang, En Zhu, Zhe Liu, Lu Zhou. Code PDF

• We remove the non-negative constraint of non-negative matrix factorization and obtain coefficient matrices with view-specific base matrices of different dimensions, then integrate them into a consensus one in a parameter-free way.

Xinhang Wan, Jiyuan Liu, Weixuan Liang, Xinwang Liu^#, Yi Wen, En Zhu. Code PDF

• We propose CMVC and it is the first attempt to handle real-time issues in late fusion multi-view clustering literature and will provide an inspiration for future research.