Neural Loop Algorithm

In this talk, I will discuss how to apply a reinforcement learning framework on the spin ice model.  Spin ice is a frustrated magnetic system with a strong topological constraint on the low-energy configurations called the ice rule. Conventional  single spin-flip Monte Carlo updates breaks this constraint.  We exploit a reinforcement learning method that parameterizes the transition operator with neural networks. By extending the Markov chain to a Markov decision process, the algorithm can adaptively search for a global update policy through its interactions with the physical model. We find that the global loop update emerges without the explicit knowledge of the ice rule. This method might serve a general framework to  search for efficient update policies in other constrained systems. 

Event Type: 
Scientific Area(s): 
Event Date: 
Mardi, Octobre 16, 2018 - 11:00 to 12:30
Sky Room
Room #: