Curvilinear Component Analysis
,Nonlinear
mapping technique developed for self-organizing neural network
,Characteristics
–More
efficient compared to other non-linear mapping algorithms: MDS-O(M3), Sammon’s mapping -O(M2N),CCA-O(M2)
–Precise
distance cost function to reduce high dimensional data with high accuracy
–Far
less problem of local minima
,MDS:
computationally intensive post-processing step
–Self-learning
capability for adding new nodes
,Not
yet explored further in our work though