Limit theorems and phase transitions in tensor versions of some common statistical physics models -II

Speaker: Sanchayan Bhowal (ISI Bangalore)

Abstract: Two classical models in statistical physics that are also used in statistics to capture dependency of discrete variables on a network structure, are the Ising and the Potts model. The classical versions of these models can incorporate pairwise dependency between the  variables only, and hence are not applicable frameworks for multi-body interactions and peer-group effects. To circumvent this issue, one can use the tensor/multi-spin versions of these models, instead. In this talk, we will discuss some of our recent results in the theory of asymptotics of the magnetization and parameter estimates in these tensor Ising and Potts models (with particular emphasis on the latter). The results unearth various surprising phase transition phenomena, such as the existence of a continuous curve in the interior of the parameter space on which the magnetization and parameter estimates have limiting mixture distributions, together with a superefficiency behavior of the parameter estimates at the boundary point(s) of this curve.

Zoom link:  https://us02web.zoom.us/j/88670406480
Meeting ID: 886 7040 6480
Passcode: No passcode