SpectralClustering

• Documentation
• Check out the Examples

The library provides functions that allow:

• Build the affinity matrix.
• Perform the embedding of the patterns in the space spanned by the eigenvectors of the matrices derived from the affinity matrix.
  • Obtain an approximation of the eigenvectors in order to reduce the computational complexity.
  • Exploiting information from multiple views. Corresponding nodes in each graph should have the same cluster membership.
• Clusterize the eigenvector space.

Methods implemented

• Graph construction
  • Self-Tuning Spectral Clustering
• Embedding
  • Normalized cuts and image segmentation
  • On Spectral Clustering: Analysis and an algorithm
  • Understanding Popout through Repulsion
  • Segmentation Given Partial Grouping Constraints
• Approximate embedding
  • Spectral grouping using the nystrom method
    • Nystrom sampling depends on the eigenspectrum shape of the data
  • Large Scale Spectral Clustering with Landmark-Based Representation
• Multiple views
  • Kernel Addition
  • Kernel Product
  • Feature Concatenation (in the examples section)
  • Co-regularized Multi-view Spectral Clustering
• Incremental
  • TODO Incremental spectral clustering by efficiently updating the eigen-system
• Clusterize
  • Multiclass Spectral Clustering
  • KMeans via Clustering.jl

The documentation and the library is still a work in progress.