Title: Cutoff for Random Walks on Contingency Tables
Speaker: Zihao Fang (OSU)
Abstract: Contingency tables are matrices with fixed row and column sums. It was of particular interest in math and statistics to sample contingency tables at random for various purposes, and one would like to know how efficient a sampling algorithm can be. We studied the Diaconis-Gangolli random walk as a Markov chain Monte Carlo model and observed cutoff phenomena, a sharp transition from almost deterministic to almost random in the context of mixing time. We gave cutoff time for the random walks on both 1*n and n*n contingency tables, and we generalized our method to show cutoff for a family of random walks on the torus (Z/qZ)^n. This is a joint work with Andrew Heeszel (OSU).
