Feature selection using Principle of Relevant Information (PRI)
Let an NxD matrix X denote feature matrix where N is the number of observations and D is the number of feature dimensionality. Our goal is to reduce the dimension of feature into a smaller set d<D using PRI.
Data set: Haxby's 8-stimuli
In the first step, rather than implementing the PRI to the data set, I use GMS, which can be viewed as a special case of PRI instead.