Multi-task self-supervised learning based fusion representation for Multi-view clustering

Tianlong Guo, Derong Shen, Yue Kou, Tiezheng Nie

Abstract: Highlights•MSLMFR takes the representation learning of different views as its mainstream task.•It shares parameters across multiple tasks to extract multi-view representation.•Three types of tasks are designed within the framework to optimize the efficiency.•Gradient is optimized by aligning matrices in orthogonal spaces and taking a norm.