Machine Learning of Implicit Combinatorial Rules in Mechanical Metamaterials

Physical Review Letters, vol. 129, pp. 198003

Abstract

Combinatorial problems arising in puzzles, origami, and (meta)material design have rare sets of solutions, which define complex and sharply delineated boundaries in configuration space. These boundaries are difficult to capture with conventional statistical and numerical methods. Here we show that convolutional neural networks can learn to recognize these boundaries for combinatorial mechanical metamaterials, down to finest detail, despite using heavily undersampled training sets, and can successfully generalize. This suggests that the network infers the underlying combinatorial rules from the sparse training set, opening up new possibilities for complex design of (meta)materials.

Authors 4

1. Ryan van Mastrigt

   Institute for Atomic and Molecular Physics · Institute of Physics · University of Amsterdam

   Affiliation as printed

   AMOLF, Science Park 104, 1098 XG Amsterdam, The Netherlands

   Institute of Physics, Universiteit van Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands

2. Marjolein Dijkstra

   Utrecht University

   Affiliation as printed

   Soft Condensed Matter, Debye Institute for Nanomaterials Science, Department of Physics, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands

3. Martin van Hecke Aachen Huygens-Kamerling Onnes Lab

   Leiden University · Institute for Atomic and Molecular Physics

   Affiliation as printed

   AMOLF, Science Park 104, 1098 XG Amsterdam, The Netherlands

   Huygens-Kamerling Onnes Lab, Universiteit Leiden, Postbus 9504, 2300 RA Leiden, The Netherlands

4. Corentin Coulais

   Institute of Physics · University of Amsterdam

   Affiliation as printed

   Institute of Physics, Universiteit van Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands