Accession Number:

AD1107752

Title:

Computationally and Statistically Efficient Model Fitting Techniques

Descriptive Note:

Technical Report

Corporate Author:

SOUTH DAKOTA STATE UNIV BROOKINGS BROOKINGS United States

Report Date:

2014-01-01

Pagination or Media Count:

7.0

Abstract:

AbstractIn large-scale stochastic simulations, analysis withsufficient accuracy is often extremely time consuming. Thecomplexity of the analysis is exacerbated with increasing dimensionalityof the parameter space and sudden abruptness inthe topology of the input-output response surface. This paperaddresses computational issues in fitting and generating errormeasures of simulation metamodels, demonstrating the merit ofhigh-performance computing in Python. We systematically comparethe speed of programming languages including MATLAB,R and Python as well as using different computing architecturesincluding high-performing laptops and high-power parallel processingclusters. The experimentation is discussed in this paperusing a simple scenario, and activities are being pursued to studyother scenarios with varying complexities that will be reported at the conference.

Subject Categories:

Distribution Statement:

APPROVED FOR PUBLIC RELEASE