StriderNet: A Graph Reinforcement Learning Approach to Optimize Atomic Structures on Rough Energy Landscapes

Vaibhav Bihani, Sahil Manchanda, Srikanth Sastry, Sayan Ranu, N M Anoop Krishnan
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:2431-2451, 2023.

Abstract

Optimization of atomic structures presents a challenging problem, due to their highly rough and non-convex energy landscape, with wide applications in the fields of drug design, materials discovery, and mechanics. Here, we present a graph reinforcement learning approach, StriderNet, that learns a policy to displace the atoms towards low energy configurations. We evaluate the performance of StriderNet on three complex atomic systems, namely, binary Lennard-Jones particles, calcium silicate hydrates gel, and disordered silicon. We show that StriderNet outperforms all classical optimization algorithms and enables the discovery of a lower energy minimum. In addition, StriderNet exhibits a higher rate of reaching minima with energies, as confirmed by the average over multiple realizations. Finally, we show that StriderNet exhibits inductivity to unseen system sizes that are an order of magnitude different from the training system. All the codes and datasets are available at https://github.com/M3RG-IITD/StriderNET.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-bihani23a, title = {{S}trider{N}et: A Graph Reinforcement Learning Approach to Optimize Atomic Structures on Rough Energy Landscapes}, author = {Bihani, Vaibhav and Manchanda, Sahil and Sastry, Srikanth and Ranu, Sayan and Krishnan, N M Anoop}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {2431--2451}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/bihani23a/bihani23a.pdf}, url = {https://proceedings.mlr.press/v202/bihani23a.html}, abstract = {Optimization of atomic structures presents a challenging problem, due to their highly rough and non-convex energy landscape, with wide applications in the fields of drug design, materials discovery, and mechanics. Here, we present a graph reinforcement learning approach, StriderNet, that learns a policy to displace the atoms towards low energy configurations. We evaluate the performance of StriderNet on three complex atomic systems, namely, binary Lennard-Jones particles, calcium silicate hydrates gel, and disordered silicon. We show that StriderNet outperforms all classical optimization algorithms and enables the discovery of a lower energy minimum. In addition, StriderNet exhibits a higher rate of reaching minima with energies, as confirmed by the average over multiple realizations. Finally, we show that StriderNet exhibits inductivity to unseen system sizes that are an order of magnitude different from the training system. All the codes and datasets are available at https://github.com/M3RG-IITD/StriderNET.} }
Endnote
%0 Conference Paper %T StriderNet: A Graph Reinforcement Learning Approach to Optimize Atomic Structures on Rough Energy Landscapes %A Vaibhav Bihani %A Sahil Manchanda %A Srikanth Sastry %A Sayan Ranu %A N M Anoop Krishnan %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-bihani23a %I PMLR %P 2431--2451 %U https://proceedings.mlr.press/v202/bihani23a.html %V 202 %X Optimization of atomic structures presents a challenging problem, due to their highly rough and non-convex energy landscape, with wide applications in the fields of drug design, materials discovery, and mechanics. Here, we present a graph reinforcement learning approach, StriderNet, that learns a policy to displace the atoms towards low energy configurations. We evaluate the performance of StriderNet on three complex atomic systems, namely, binary Lennard-Jones particles, calcium silicate hydrates gel, and disordered silicon. We show that StriderNet outperforms all classical optimization algorithms and enables the discovery of a lower energy minimum. In addition, StriderNet exhibits a higher rate of reaching minima with energies, as confirmed by the average over multiple realizations. Finally, we show that StriderNet exhibits inductivity to unseen system sizes that are an order of magnitude different from the training system. All the codes and datasets are available at https://github.com/M3RG-IITD/StriderNET.
APA
Bihani, V., Manchanda, S., Sastry, S., Ranu, S. & Krishnan, N.M.A.. (2023). StriderNet: A Graph Reinforcement Learning Approach to Optimize Atomic Structures on Rough Energy Landscapes. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:2431-2451 Available from https://proceedings.mlr.press/v202/bihani23a.html.

Related Material