[AIRG] Correlation explanation, Nov. 28

[Lucas Morton <lamorton@xxxxxxxx>]

Is there a principled way to learn the underlying structure of a data source in an unsupervised manner, with minimal prior knowledge? A promising approach appeals to Reichenbach's principle: every correlation deserves an explanation. Formalizing this intuition leads to a general information-theoretic approach to unsupervised learning.

Discovering Structure in High-Dimensional Data Through Correlation Explanation

Greg Ver Steeg, Aram Galstyan

NIPS 2014

https://arxiv.org/abs/1406.1222

4pm, Wednesday November 28

CS 4310

--Lucas

--------------------------

Lucas A. Morton

Postdoctoral Researcher

Plasma Spectroscopy for Turbulence and Instabilities Group

Department of Engineering Physics

UW - Madison