Automating Science: Anti-streams, Universal Similarity and Statistical Causality

Event Sponsor: 
Mathematics and Computing Science Seminar
Start Date: 
May 4 2015 - 10:30am
Building 240/Room 1406-1407
Argonne National Laboratory
Ishanu Chattopadhyay
Speaker(s) Title: 
University of Chicago
Marc Snir

Despite prevalence of computational tools, the fundamental step of making new hypotheses has always been ultimately driven by human insight.  It has been the human scientist at the center stage - doing the “science”, with machines assisting by carrying out routine calculations. The notion of automating scientific discovery is based on the possibility of reversing these roles. In this talk, I will attempt to make a case in the light of new breakthroughs in automated reasoning, zero-knowledge inference, and the new computable metrics for universal similarity and statistical causality.  I show how machine learning may be carried out without choosing features, how we can carry out non-parametric analysis in the absence of prior knowledge and yet infer generative models, and how we can go beyond computing correlations and begin computing causation in data.

Miscellaneous Information: 

Biographical Sketch:

Dr. Chattopadhyay is a Research Scientist with the Computation Institute, and the Institute for Genomics and Systems Biology at the University of Chicago.  His research focuses on the theory of unsupervised machine learning, the interplay of stochastic processes and formal language theory in exploring the limits of automated learning algorithms, and the philosophical and mathematical underpinnings of the question of inferring causality from data.  His most visible contributions include the algorithms for data smashing, inverse Gillespie inference, and non-parametric implements of Granger causal inference.