A Simple Algorithm for Scalable Monte Carlo Inference

Speaker: Maksym Byshkin, Università della Svizzera italiana

Date: Tuesday, February 19, 2019

Place: USI Lugano Campus, room SI-003, Informatics building (Via G. Buffi 13)

Time: 12:30-13:30

Abstract:

Statistical inference involves estimation of parameters of a model based on observations. Building on the recently proposed Equilibrium Expectation approach and Persistent Contrastive Divergence, we derive a simple and fast Markov chain Monte Carlo algorithm for maximum likelihood estimation (MLE) of parameters of exponential family distributions. The algorithm has good scaling properties and is suitable for Monte Carlo inference on large network data with billions of tie variables. The performance of the algorithm is demonstrated on Markov random fields, conditional random fields, exponential random graph models and Boltzmann machines.

Biography:

Byshkin Maksym is a Swiss National Science Foundation postdoctoral research fellow in the Social Network Analysis Research Centre (SoNAR-C) at the Institute of Computational Sciences (ICS) and the InterDisciplinary Institute of Data Science (IDIDS), Università della Svizzera italiana (USI). His research fields are statistical physics, statistical network analysis, computational chemistry and interdisciplinary collaborations. The research activity and research interests are focused on developments of empirical models and computational methods for high performance computing.

Host: Prof. Ernst Wit