Date of Award

Spring 4-24-2019

Document Type

Honors Project

Degree Name

Bachelor of Science

Department

Mathematics

Department Chair or Program Director

Helmstutler, Randall

First Advisor

Collins, Jeb

Major or Concentration

Mathematics

Abstract

The purpose of this research is to use machine learning algorithms to predict the fastest settings for a program called Bertini. Bertini is a computer program that approximates solutions to systems of polynomial equations. Settings for this program can be changed by the user, but the user may not necessarily know the best settings to use to optimize the run time for a particular system of polynomial equations. The settings that were focused on were the differential equation predictor methods when tracking the homotopy to the solution of the system. A neural network was used on a training set of data to create a model and then a test set was run through to obtain a percent accuracy for this model. Increased accuracy for the model was obtained by changing hyperparameters of the neural network. Neural networks with training sets of 3,000 and 8,000 polynomials were used and results were found for different parameter settings.

Included in

Mathematics Commons

Share

COinS