Hierarchical Clustering
-- Decision Tree
Partitioning Clustering
-- k-means
Density-based Clustering
-- EM algorithm
Topology-based Clustering
-- Self-organizing Map
(SOM)
-- Growing Neural Gas
(GNG)
[1]R.Chalasani and J.C.Principe. "Self-organizing maps with information theorelic learning, "Neurocomputing, vol. 147, pp. 3-14,2015.
where, k denotes the number of nodes in network.W_{s} denotes the reference vector, and σ denotes a kernel bandwidth.
Err ← Err+[CIM(v(l),w)]_{2}
:learning rale
where, G_{σ} and h_{σ}^{es} are as follows;
here, σ in G_{σ} and h_{σ}^{es} are both same kernel bandwidth.
where, Δσ(l) is calculated as follow;
here, ρ denotes a scaling factor for the stability of topological network [3].
GNG-CIM with several kernel bandwidth condition
Original GNG
GNG-CIM (Num. of Nodes)
GNG-CIM (Node Distribution)
where