The paper introduces FMMVCC, an unsupervised deep-clustering framework for univariate time series. It combines Mamba state-space sequence modeling with fuzzy clustering and multi-view self-supervised learning based on temporal masking and augmentations. The authors state that the method captures long-range dependencies with linear complexity while avoiding the higher computational cost of some existing architectures. Across 15 benchmark datasets and 60 metric evaluations, FMMVCC reportedly achieves the best result in 29 evaluations and the highest average rank across all tested scenarios.
No heat snapshots are available in the last 24 hours.