Learning Potential Energy Landscapes using Graph Kernels
By Grégoire Ferré, Terry S. Haut, Kipton Barros. 2016
Recent machine learning methods make it possible to model potential energy of atomic configurations with chemical-level accuracy (as calculated from ab-initio calculations) and at speeds suitable for molecular dynamics simulation. Best performance is achieved when the known physical constraints are encoded in the machine learning models. For example, the atomic energy is invariant under global translations and rotations; it is also invariant to permutations of same-species atoms. Although simple to state, these symmetries are complicated to encode into machine learning algorithms. In this paper, we present a machine learning approach based on graph theory that naturally incorporates translation, rotation, and permutation symmetries. Specifically, we use a random walk graph kernel to measure the similarity of two adjacency matrices, each of which represents a local atomic environment. We show on a standard benchmark that our Graph Approximated Energy (GRAPE) method is competitive with state of the art kernel methods. Furthermore, the GRAPE framework is flexible and admits many possible extensions.
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Stefan Chmiela
- K. Müller
- Wojciech Samek
- Alexandre Tkatchenko
- Grgoire Montavon
- Huziel E. Sauceda
- Oliver T. Unke
- Kristof T. Schütt
Citations in the corpus
- Towards exact molecular dynamics simulations with machine-learned force fields (2018, methodology): “We start with adjacency matrices as representation for the molecular graph [9, 13, 47, 53, 54].”