Towards an Analytical Definition of Sufficient Data

Abstract

We show that, for each of five datasets of increasing complexity, certain training samples are more informative of class membership than others. These samples can be identified a priori to training by analyzing their position in reduced dimensional space relative to the classes’ centroids. Specifically, we demonstrate that for the datasets studied, samples that are nearer to their classes’ centroids are less informative than those that are furthest from them. For all five datasets studied, we show that there is no statistically significant difference between training on the entire training set and when excluding up to 2% of the data nearest to each class’s centroid.