Abstract:

A duality arises when a given physical system has two different mathematical representations. In this talk I will discuss the possibility of using modern machine learning methods to find dualities in statistical physics. Establishing a duality in lattice statistical mechanics models requires the construction of a dual Hamiltonian and a map from the original to the dual observables. By using simple neural networks to parameterize these maps and introducing a loss function that penalises the difference between correlation functions in original and dual models, the process of duality discovery can be formulated as an optimization problem. I will introduce the required concepts from machine learning and show how to solve this problem numerically for the 2d Ising model and some variants. I will also discuss the prospects of finding new dualities using such methods.

Zoom-Link:

https://univienna.zoom.us/j/63690100486?pwd=Sh7PzkBqgQZCcNOLseQFRqOmtMgnpp.1

Meeting ID: 636 9010 0486
Passcode: 211228