Machine learning that predicts well may not learn the correct physical descriptions of glassy systems
arXiv:2403.00751 · doi:10.1103/PhysRevResearch.6.033091
Abstract
The complexity of glasses makes it challenging to explain their dynamics. Machine Learning (ML) has emerged as a promising pathway for understanding glassy dynamics by linking their structural features to rearrangement dynamics. Support Vector Machine (SVM) was one of the first methods used to detect such correlations. Specifically, a certain output of SVMs trained to predict dynamics from structure, the distance from the separating hyperplane, was interpreted as being linearly related to the activation energy for the rearrangement. By numerical analysis of toy models, we explore under which conditions it is possible to infer the energy barrier to rearrangements from the distance to the separating hyperplane. We observe that such successful inference is possible only under very restricted conditions. Typical tests, such as the apparent Arrhenius dependence of the probability of rearrangement on the inferred energy and the temperature, or high cross-validation accuracy do not guarantee success. We propose practical approaches for measuring the quality of the energy inference and for modifying the inferred model to improve the inference, which should be usable in the context of realistic datasets.
11 pages, 7 figures
References in corpus (25)
- Machine learning and the physical sciences
- Machine learning phases of matter
- Dynamical heterogeneities in a supercooled Lennard-Jones liquid
- Learning phase transitions by confusion
- Discovering Phase Transitions with Unsupervised Learning
- Identifying structural flow defects in disordered solids using machine learning methods
- Perspective: The Glass Transition
- Discovering Phases, Phase Transitions and Crossovers through Unsupervised Machine Learning: A critical examination
- The Principles of Deep Learning Theory
- The Relationship Between Local Structure and Relaxation in Out-of-Equilibrium Glassy Systems
- Disconnecting structure and dynamics in glassy thin films
- Averaging local structure to predict the dynamic propensity in supercooled liquids
- Deep Learning Topological Invariants of Band Insulators
- Predicting dynamic heterogeneity in glass-forming liquids by physics-inspired machine learning
- Nuclear liquid-gas phase transition with machine learning
- Machine Learning as a universal tool for quantitative investigations of phase transition
- Machine Learning Characterization of Structural Defects in Amorphous Packings of Dimers and Ellipses
- Improving the prediction of glassy dynamics by pinpointing the local cage
- Identification of hidden order and emergent constraints in frustrated magnets using tensorial kernel methods
- Signature of Dynamical Heterogeneity in Spatial Correlations of Particle Displacement and its Temporal Evolution in Supercooled Liquids
- Dynamic heterogeneity at the experimental glass transition predicted by transferable machine learning
- Geometry-enhanced graph neural network for learning the smoothness of glassy dynamics from static structure
- Rotation-equivariant Graph Neural Networks for Learning Glassy Liquids Representations
- Machine Learning Phase Transition: An Iterative Proposal
- Building a "trap model" of glassy dynamics from a local structural predictor of rearrangements