Optimization and Supervised Machine Learning Methods for Fitting Numerical Physics Models without Derivatives

Abstract

We address the calibration of a computationally expensive nuclear physics model for which derivative information with respect to the fit parameters is not readily available. Of particular interest is the performance of optimization-based training algorithms when dozens, rather than millions or more, of training data are available and when the expense of the model places limitations on the number of concurrent model evaluations that can be performed. As a case study, we consider the Fayans energy density functional model, which has characteristics similar to many model fitting and calibration problems in nuclear physics. We analyze hyperparameter tuning considerations and variability associated with stochastic optimization algorithms and illustrate considerations for tuning in different computational settings.

arXiv comment25-page article, 9-page supplement, 1-page notice

References in corpus (19)

• Nuclear Energy Density Optimization
• Nuclear energy density optimization: Large deformations
• Derivative-free optimization methods
• An optimized chiral nucleon-nucleon interaction at next-to-next-to-leading order
• Error Estimates of Theoretical Models: a Guide
• Nuclear energy density optimization: Shell structure
• Local chiral potentials and the structure of light nuclei
• An Optimal Algorithm for Bandit and Zero-Order Convex Optimization with Two-Point Feedback
• Measurement and microscopic description of odd-even staggering of charge radii of exotic copper isotopes
• Uncertainty Quantification for Nuclear Density Functional Theory and Information Content of New Measurements
• Towards the global description of nuclear charge radii: exploring the Fayans energy density functional
• Spurious finite-size instabilities in nuclear energy density functionals
• Beyond the charge radius: the information content of the fourth radial moment
• The importance of better models in stochastic optimization
• A Bayesian Approach for Parameter Estimation and Prediction using a Computationally Intensive Model
• Statistical uncertainties of a chiral interaction at next-to-next-to leading order
• Derivative-free optimization for parameter estimation in computational nuclear physics
• Bayesian optimization in ab initio nuclear physics
• Adaptive Sampling Quasi-Newton Methods for Zeroth-Order Stochastic Optimization

Cited by in corpus (7)

• Machine Learning in Nuclear Physics
• Nuclear Charge Radii of the Nickel Isotopes ${}^{58-68,70}$Ni
• Skyrme-Hartree-Fock-Bogoliubov mass models on a 3D Mesh: Effect of triaxial shape
• Improved description of nuclear charge radii: Global trends beyond $N=28$ shell closure
• A Bayesian mixture model approach to quantifying the empirical nuclear saturation point
• Constructing a simulation surrogate with partially observed output
• Designing a Framework for Solving Multiobjective Simulation Optimization Problems