Machine Learning for High-Dimensional Dynamic Stochastic Economies

38 Pages Posted: 6 Mar 2017 Last revised: 10 Aug 2017

See all articles by Simon Scheidegger

Simon Scheidegger

University of Lausanne - School of Economics and Business Administration (HEC-Lausanne)

Ilias Bilionis

Purdue University

Date Written: June 23, 2017

Abstract

We present the first computational framework that can compute global solutions to very-high-dimensional dynamic stochastic economic models on arbitrary state space geometries.

This framework can also resolve value and policy functions' local features and perform uncertainty quantification, in a self-consistent manner. We achieve this by combining Gaussian process machine learning with the active subspace method; we then embed this into a massively parallelized discrete-time dynamic programming algorithm. To demonstrate the broad applicability of our method, we compute solutions to stochastic optimal growth models of up to 500 continuous dimensions. We also show that our framework can address parameter uncertainty and can provide predictive confidence intervals for policies that correspond to the epistemic uncertainty induced by limited data. Finally, we propose an algorithm, based on this framework, that is capable of learning irregularly shaped ergodic sets as well as performing dynamic programming on them.

Keywords: Machine Learning, Gaussian Process, Active Subspaces, High-Performance Computing, Neoclassical Growth, Dynamic Programming

JEL Classification: C63, C68, E130

Suggested Citation

Scheidegger, Simon and Bilionis, Ilias, Machine Learning for High-Dimensional Dynamic Stochastic Economies (June 23, 2017). Available at SSRN: https://ssrn.com/abstract=2927400 or http://dx.doi.org/10.2139/ssrn.2927400

Simon Scheidegger (Contact Author)

University of Lausanne - School of Economics and Business Administration (HEC-Lausanne) ( email )

Unil Dorigny, Batiment Internef
Lausanne, 1015
Switzerland

Ilias Bilionis

Purdue University ( email )

610 Purdue Mall
West Lafayette, IN 47907
United States

Register to save articles to
your library

Register

Paper statistics

Downloads
621
rank
40,936
Abstract Views
2,147
PlumX Metrics