FPMVS-CAG

FPMVS-CAG performs multi-view subspace clustering by integrating anchor selection and subspace graph construction into a unified optimization to produce parameter-free, scalable anchor subspace graphs for fusing multi-view data.

Key Features:

• Integrated optimization: Integrates anchor selection with subspace graph construction in a unified optimization framework that allows both processes to mutually inform each other.
• Anchor sampling mechanism: Identifies landmark anchors via an anchor sampling mechanism to represent the dataset more effectively than heuristic methods such as k-means or uniform sampling.
• Linear time complexity: Achieves linear time complexity with respect to sample size, addressing the cubic time complexity of traditional subspace clustering approaches.
• Parameter-free learning: Automatically learns an optimal anchor subspace graph without requiring hyper-parameter tuning.
• Validated efficiency and effectiveness: Demonstrates improved computational efficiency and clustering effectiveness on multiple benchmark datasets for large-scale applications.

Scientific Applications:

• Large-scale multi-view subspace clustering: Applicable to clustering tasks involving many samples across multiple complementary views.
• Multi-view data fusion in bioinformatics: Constructs anchor-based subspace graphs to fuse multi-view information for downstream analysis in bioinformatics.

Methodology:

Integrates anchor selection and subspace graph construction into a unified optimization; employs an anchor sampling mechanism to identify landmarks; automatically learns an optimal anchor subspace graph and attains linear time complexity with respect to sample size.