authorfiveaioxford-research-groupedinburghspectral-clusteringdirected-graph

Steinar Laenen

Affiliation: Oxford Research Group, FiveAI; previously MSc student at the University of Edinburgh (during the work of entities/Laenen Sun 2020 Higher-Order Spectral Clustering).

Email: steinar.laenen@five.ai / steinar9@gmail.com

Contributions to this wiki

  • Lead author of entities/Laenen Sun 2020 Higher-Order Spectral Clustering (NeurIPS 2020) — first paper providing Cheeger-style theoretical guarantees for directed-graph spectral clustering under general (non-DSBM) settings, via the Hermitian Laplacian L_G = I - D^{-1/2} A D^{-1/2} with k-th root of unity phase encoding.

The Hermitian Laplacian framework has been adopted as an alternative to entities/Mihai Cucuringu's Hermitian adjacency in financial lead-lag applications (entities/Bennett Cucuringu Reinert 2022 Lead-Lag Network Clustering §3.3 cites this as an alternative).

Research themes

  • Directed graph clustering with theoretical guarantees
  • Spectral sparsification for nearly-linear time graph algorithms
  • Higher-order network structure (ordered chain topology in clusters)

See also