Tight clusters and smooth manifolds with the harmonic topographic map
Marian Peña, Colin Fyfe · SMO'05 Proceedings of the 5th WSEAS international conference on Simulation, modelling and optimization · 2005
We review a new form of self-organizing map introduced in [5] which is based on a non-linear projection of latent points into data space, identical to that performed in the Generative Topographic Mapping (GTM) [1]. We discuss a refinement of that mapping (M-HaToM) and show on real and artificial data how it both finds the true manifold on which a data set lies and also clusters data more tightly than the previous algorithm (D-HaToM).