Method of learning conditional probability distributions, uses partition of the input space into regions where local subzone output vectors are broadly the same as the overall output and uses statistical testing of distributions

摘要

The learning method partitions the input vector space into regions, where each region is such that the conditional probability distribution of the output for all vectors in this region is not significantly different to distribution of outputs in this region, and each region contains a subset of learning vectors sufficiently large to allow the application of a statistical test to the distribution.