6.986 Inference and Information (H)

L TR9:30-11, Room 2-105
Professor Greg Wornell, gww@mit.edu, Professor Polina Golland, polina@csail.mit.edu, Professor Alan Willsky, willsky@mit.edu
Prereq.: 6.041/6.431 or 6.436 or 6.011
4-0-8

This subject qualifies as a Communication, Control and Signal Processing Engineering Concentration subject.

Introduction to principles of statistical inference. Models, likelihoods, beliefs. Hypothesis testing; parameter estimation. Sufficient statistics; exponential families. Kullback-Leibler divergence, entropy, mutual information and Fisher information in inference. Selection of priors; model capacity. Approximate marginalization and inference; Monte-Carlo methods. Large-sample asymptotics; error exponents. Parametric modeling; order estimation; nonparametric modeling; clustering. Selected special topics.

NOTE: This subject replaces, in part, the now-retired 6.432 (Stochastic Processes, Detection, and Estimation). Those having taken 6.432 for credit may also take 6.986 for credit, but petitions to use both 6.432 and 6.986 for TQE credit will not be approved.