Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Multiobjective Hyperparameter Optimization for Deep Learning Interatomic Potential Training Using NSGA-II

Conference ·

Deep neural network (DNN) potentials are an emerging tool for simulation of dynamical atomistic systems, with the promise of quantum mechanical accuracy at speedups of 10000$$\times$$. As with other DNN methods, hyperparameters used during training can make a substantial difference in model accuracy, and optimal settings vary with dataset. To enable rapid tuning of hyperparameters for DNN potential training, we developed a scalable multiobjective optimization evolutionary algorithm for supercomputers and tested it on the Summit system at the Oak Ridge Leadership Computing Facility (OLCF). The multiobjective approach is required due to the coupling of two learned values defining the potential: the energy and force. Using a large-scale implementation of the NSGA-II algorithm adapted for training DNN potentials, we discovered several optimal multiobjective combinations, including best choices of activation functions, learning rate scaling scheme, and pairing of the two radial cutoffs used in the three dimensional descriptor function.

Research Organization:
Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States)
Sponsoring Organization:
USDOE
DOE Contract Number:
AC05-00OR22725
OSTI ID:
1996670
Country of Publication:
United States
Language:
English

References (17)

High-Performance Molecular Dynamics Simulation for Biological and Materials Sciences: Challenges of Performance Portability conference November 2018
Global Partitioning Elevation Normalization Applied to Building Footprint Prediction journal January 2020
A Holistic Approach for Elucidating Local Structure, Dynamics, and Speciation in Molten Salts with High Structural Disorder journal September 2021
Library for evolutionary algorithms in Python (LEAP)
  • Coletti, Mark A.; Scott, Eric O.; Bassett, Jeffrey K.
  • GECCO '20: Genetic and Evolutionary Computation Conference, Proceedings of the 2020 Genetic and Evolutionary Computation Conference Companion https://doi.org/10.1145/3377929.3398147
conference July 2020
DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics journal July 2018
mpi4py: Status Update After 12 Years of Development journal July 2021
On-the-fly machine learning force field generation: Application to melting points journal July 2019
An Empirical Study of the Impact of Hyperparameter Tuning and Model Optimization on the Performance Properties of Deep Neural Networks journal April 2022
E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials journal May 2022
Automating Configuration of Convolutional Neural Network Hyperparameters Using Genetic Algorithm journal January 2020
Hyper-parameter optimization for convolutional neural network committees based on evolutionary algorithms conference September 2017
Machine Learning Force Fields journal March 2021
An improved hyperparameter optimization framework for AutoML systems using evolutionary algorithms journal March 2023
Development and Validation of a DFT-Based Force Field for a Hydrated Homoalanine Polypeptide journal February 2021
CP2K: An electronic structure and molecular dynamics software package - Quickstep: Efficient and accurate electronic structure calculations journal May 2020
Evolving Larger Convolutional Layer Kernel Sizes for a Settlement Detection Deep-Learner on Summit conference November 2019
Diagnosing autonomous vehicle driving criteria with an adversarial evolutionary algorithm
  • Coletti, Mark A.; Gao, Shang; Paulissen, Spencer
  • GECCO '21: Genetic and Evolutionary Computation Conference, Proceedings of the Genetic and Evolutionary Computation Conference Companion https://doi.org/10.1145/3449726.3459573
conference July 2021

Similar Records

Streamlining Ocean Dynamics Modeling with Fourier Neural Operators: A Multiobjective Hyperparameter and Architecture Optimization Approach
Journal Article · Fri May 10 00:00:00 EDT 2024 · Mathematics · OSTI ID:2477212

Improving Deep Neural Networks’ Training for Image Classification With Nonlinear Conjugate Gradient-Style Adaptive Momentum
Journal Article · Fri Mar 24 00:00:00 EDT 2023 · IEEE Transactions on Neural Networks and Learning Systems · OSTI ID:2280651

NSGA-PINN: A Multi-Objective Optimization Method for Physics-Informed Neural Network Training
Journal Article · Mon Apr 03 00:00:00 EDT 2023 · Algorithms · OSTI ID:1968107

Related Subjects