12
9

Separating populations with wide data: A spectral analysis

Abstract

In this paper, we consider the problem of partitioning a small data sample drawn from a mixture of kk product distributions. We are interested in the case that individual features are of low average quality γ\gamma, and we want to use as few of them as possible to correctly partition the sample. We analyze a spectral technique that is able to approximately optimize the total data size--the product of number of data points nn and the number of features KK--needed to correctly perform this partitioning as a function of 1/γ1/\gamma for K>nK>n. Our goal is motivated by an application in clustering individuals according to their population of origin using markers, when the divergence between any two of the populations is small.

View on arXiv
Comments on this paper