Machine-learning-assisted Monte Carlo fails at sampling computationally hard problems. Issue 1 (14th March 2023)

• Record Type: Journal Article
• Authors: Ciarella, Simone; Trinquier, Jeanne; Weigt, Martin; Zamponi, Francesco
• Is Part Of: Machine learning: science and technology. Volume 4:Issue 1(2023)
• Publication Date: 2023-03-14
• Subjects: statistical physics -- Monte Carlo Markov Chains -- sampling; 006.31
• DOI: 10.1088/2632-2153/acbe91
• Languages: English
• ISSNs: 2632-2153

Abstract

Abstract: Several strategies have been recently proposed in order to improve Monte Carlo sampling efficiency using machine learning tools. Here, we challenge these methods by considering a class of problems that are known to be exponentially hard to sample using conventional local Monte Carlo at low enough temperatures. In particular, we study the antiferromagnetic Potts model on a random graph, which reduces to the coloring of random graphs at zero temperature. We test several machine-learning-assisted Monte Carlo approaches, and we find that they all fail. Our work thus provides good benchmarks for future proposals for smart sampling algorithms.