E. López-Rubio, J.M. Ortiz-de-Lazcano-Lobato, and M. del Carmen Vargas-González (Spain)
Neural networks, local PCA, competitive learning
One of the best known techniques for multidimensional data analysis is the Principal Components Analysis (PCA). A number of local PCA neural models have been proposed to partition an input distribution into meaningful clusters where each neuron uses a certain number of basis vectors to represent the principal directions of a particular cluster. Most of these neural networks are unable to learn the number of basis vectors, which is specified a priori as a fixed parameter. We study several approaches to infer the appropriate dimensionality of the clusters when we combine Probabilistic Principal Components Analysis Neurons in a competitive way. Finally, experimental results are presented to show the performance of our networks outperforms the mixture of gaussian models presented by Tipping and Bishop.
Important Links:
Go Back
