SPUDS: Spectral Clustering Algorithm with Parameter Selection Recommendations

An R implementation of a spectral clustering algorithm proposed by David P. Hofmeyr. The algorithm addresses the practical challenges of selecting the number of clusters and tuning the scaling parameter by leveraging the asymptotic value of the normalised cut. It is available as an open-source package on GitHub.

Use Cases

• Clustering high-dimensional data based on the spectral clustering method described.
• Selecting the number of clusters in a data-driven manner based on the asymptotic analysis of the normalised cut.
• Tuning the scaling parameter for spectral clustering based on connections to density clustering identified in the paper.
• Benchmarking clustering algorithm performance in applied domains as referenced in the description.

Strengths

• Algorithm is based on theoretical analysis of the normalised cut's asymptotic value.
• Includes specific recommendations for selecting the number of clusters and the scaling parameter.
• Open-source R implementation is publicly available for verification and use.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

David P. Hofmeyr

Collection Method

Algorithmic research and implementation.