Identifying the relevant degrees of freedom in a complex physical system is a key stage in developing effective theories in and out of equilibrium. The celebrated renormalization group provides a framework for this, but its practical execution in unfamiliar systems is fraught with ad hoc choices, whereas machine learning approaches, though promising, lack formal interpretability. Here we present an algorithm employing state-of-the-art results in machine-learning-based estimation of information-theoretic quantities, overcoming these challenges, and use this advance to develop a new paradigm in identifying the most relevant operators describing properties of the system. We demonstrate this on an interacting model, where the emergent degrees of freedom are qualitatively different from the microscopic constituents. Our results push the boundary of formally interpretable applications of machine learning, conceptually paving the way toward automated theory building.
Bibliographical noteFunding Information:
M. K.-J. is grateful to F. Alet for his comments on the physics of the interacting dimer model. D. E. G., S. D. H., and M. K.-J. gratefully acknowledge financial support from the Swiss National Science Foundation and the NCCR QSIT, and the European Research Council under the Grant Agreement No. 771503 (TopMechMat), as well as from European Union’s Horizon 2020 Programme under Marie Sklodowska-Curie Grant Agreement No. 896004 (COMPLEX ML). Z. R. acknowledges support from ISF Grant No. 2250/19. Some of the computations were performed using the Leonhard cluster at ETH Zurich. This work was supported by a grant from the Swiss National Supercomputing Centre (CSCS) under project ID eth5b.
© 2021 American Physical Society.