An implementation of the algorithm proposed by Daniele Casagrande and Alessandro Astolfi in their 2008 paper "A Hamiltonian-Based Algorithm for Measurements Clustering". The clustering function is regarded as a Hamiltonian function, and the level lines are determined as trajectories of the associated system.
Stage 1: Basic implementation
- simple numerical methods to solve trajectories and determine cluster validity
- manual tuning of parameters
Future developments:
- automate tuning
- more sophisticated methods for solving systems of differential equations, numerical integration
- adapt our implementation to match the refinements in Casagrande, Astolfi and Sassano's 2012 paper on Hamiltonian clustering
Thanks to Grisha Ivanov for finding the paper!