On Bounding and Approximating Functions of Multiple Expectations using Quasi-Monte Carlo

TL;DR

This paper extends Monte Carlo and Quasi-Monte Carlo methods to efficiently approximate functions of multiple expectations with adaptive sampling, supporting general error criteria and implemented in a Python package for applications in machine learning and sensitivity analysis.

Contribution

It introduces a novel extension of QMC methods for functions of multiple expectations with adaptive sampling and general error criteria, implemented in QMCPy.

Findings

Supports adaptive sampling for complex functions of expectations

Implemented in an open-source Python package

Demonstrated on machine learning and sensitivity analysis problems

Abstract

Monte Carlo and Quasi-Monte Carlo methods present a convenient approach for approximating the expected value of a random variable. Algorithms exist to adaptively sample the random variable until a user defined absolute error tolerance is satisfied with high probability. This work describes an extension of such methods which supports adaptive sampling to satisfy general error criteria for functions of a common array of expectations. Although several functions involving multiple expectations are being evaluated, only one random sequence is required, albeit sometimes of larger dimension than the underlying randomness. These enhanced Monte Carlo and Quasi-Monte Carlo algorithms are implemented in the QMCPy Python package with support for economic and parallel function evaluation. We exemplify these capabilities on problems from machine learning and global sensitivity analysis.