Accelerating hybrid stochastic algorithm for spin models

Elbio R. Dagotto, John B. Kogut

Giuseppe Scriva, Emanuele Costa, Benjamin McNaughton +1 more

Adiabatic quantum computers, such as the quantum annealers commercialized by D-Wave Systems Inc., are routinely used to tackle combinatorial optimization problems. In this article, we show how to exploit them to accelerate equilibrium Markov chain Monte Carlo simulations of computationally challenging spin-glass models at low but finite temperatures. This is achieved by training generative neural networks on data produced by a D-Wave quantum annealer, and then using them to generate smart proposals for the Metropolis-Hastings algorithm. In particular, we explore hybrid schemes by combining single spin-flip and neural proposals, as well as D-Wave and classical Monte Carlo training data. The hybrid algorithm outperforms the single spin-flip Metropolis-Hastings algorithm. It is competitive with parallel tempering in terms of correlation times, with the significant benefit of a much shorter equilibration time.

Harold Falk

S. Ramasesha

I G Blagoveshchenskiy, Anna A. Ilmushkina, Vladislav E. Vilkov +1 more

This article addresses the problem of constructing a distributed task management model in a hybrid quantum-classical computing environment. The relevance of this topic stems from the need for efficient distribution of computing resources in the face of limited availability of quantum devices and the high complexity of applied problems requiring quantum acceleration. A three-tier architecture of the hybrid system is proposed, with responsibilities divided between the task, control, and hardware infrastructure levels. Dispatching and adaptive scheduling principles are described, taking into account the specific features of variational quantum-classical algorithms, the stochastic nature of quantum computing, and the capacity limitations of quantum resources. A methodology for formalizing hybrid workflows, selecting optimality criteria, and synthesizing control rules is developed based on queuing theory, Markov processes, and multi-agent models.

Yutaka Okabe, Yusuke Tomita, Chiaki Yamaguchi