Work

Gaussian processes for emulating chiral effective field theory describing few-nucleon systems

Year: 2017
Type: dissertation
Abstract: Gaussian processes (GPs) can be used for statistical regression, i.e. to predict new data given a set of observed\ndata. In this context, we construct GPs to emulate the calculation of low energy proton-neutron scattering\ncross sections and the binding energy of the helium-4 nucleus. The GP regression uses so-called kernel functions\nto approximate the covariance between observed and unknown data points. The emulation is done in an\nattempt to reduce the large computational cost associated with exact numerical simulation of the observables.\nThe underlying physical theory of the simulation is EFT. This theory enables a perturbative description of\nlow-energy nuclear forces and is governed by a set of low-energy constants to define the terms in the effective\nLagrangian. We use the research code nsopt to simulate selected observables using EFT.\nThe GPs used in this thesis are implemented using the Python framework GPy. To measure the performance\nof a GP we define an error measure called model error by comparing exact simulations to emulated predictions.\nWe also study the time and memory consumption of GPs. The choice of input training data affects the\npredictive accuracy of the resulting GP. Therefore, we examined different sampling methods with varying\namounts of data.\nWe found that GPs can serve as an effective and versatile approach for emulating the examined observables.\nAfter the initial high computational cost of training, making predictions with GPs is quick. When trained using\nthe right methods, they can also achieve high accuracy. We concluded that the Matérn 5/2 and RBF kernels\nperform best for the observables studied. When sampling input points in high dimensions, latin hypercube\nsampling is shown to be a good method. In general, with a multidimensional input space, it is a good choice to\nuse a kernel function with different sensitivities in different directions. When working with data that spans\nover many orders of magnitude, logarithmizing the data before training also improves the GP performance.\nGPs do not appear to be a suitable method for making extrapolations from a given training set, but performs\nwell with interpolations.
Language: English
Open Access status: green
OpenAlex ID: W2903553124
Cites: 10
Cited by:
Related to: 10