Extrapolating from neural network models: a cautionary tale

Abstract

We present three different methods to estimate error bars on the predictions made using a neural network. All of them represent lower bounds for the extrapolation errors. For example, we did not include an analysis on robustness against small perturbations of the input data. At first, we illustrate the methods through a simple toy model, then, we apply them to some realistic cases related to nuclear masses. By using theoretical data simulated either with a liquid-drop model or a Skyrme energy density functional, we benchmark the extrapolation performance of the neural network in regions of the Segrè chart far away from the ones used for the training and validation. Finally, we discuss how error bars can help identifying when the extrapolation becomes too uncertain and thus unreliable

References in corpus (18)

• Physics of Neutron Star Crusts
• Microscopic mass formulae
• Nuclear Energy Density Optimization
• Error Estimates of Theoretical Models: a Guide
• Nuclear Mass Predictions for the Crustal Composition of Neutron Stars: A Bayesian Neural Network Approach
• Local Density Approximation for proton-neutron pairing correlations. I. Formalism
• Nuclear energy density optimization: Shell structure
• Bayesian approach to model-based extrapolation of nuclear observables
• Refining mass formulas for astrophysical applications: a Bayesian neural network approach
• Nuclear mass systematics using neural networks
• Taming nuclear complexity with a committee of multilayer neural networks
• An artificial neural network application on nuclear charge radii
• Covariance analysis for Energy Density Functionals and instabilities
• Solution of Hartree-Fock-Bogoliubov equations and fitting procedure using N2LO Skyrme pseudo-potential in spherical symmetry
• Large-Scale Self-Consistent Nuclear Mass Calculations
• Impact of statistical uncertainties on the composition of the outer crust of a neutron star
• Trees and Forests in Nuclear Physics
• An introduction to Bootstrap for nuclear physics

Cited by in corpus (5)

• Machine Learning in Nuclear Physics
• Physically Interpretable Machine Learning for nuclear masses
• Application of multilayer perceptron with data augmentation in nuclear physics
• Controlling extrapolations of nuclear properties with feature selection
• From nuclei to neutron stars: simple binding energy computer modelling in the classroom (Part 1)