Solving Boltzmann optimization problems with deep learning

This paper presents a novel machine learning approach for efficiently solving Boltzmann probability optimization problems, which are crucial for designing efficient Ising-based hardware. The approach combines deep neural networks and random forests to minimize errors in the Ising model, addressing the challenges posed by the computational intractability of calculating the Boltzmann distribution. The authors also detail a process to express a Boltzmann probability optimization problem as a supervised machine learning problem and analyze the performance improvements over traditional optimization methods.

Fiona Knoll, John Daly, Jess Meyer