Event Title

Defining Complexity of Continuous Functions in Relation to Artificial Neural Networks

Faculty Sponsor(s)

Emma Wright

Abstract

Artificial Neural Networks are general function approximators. Some architectures of an Artificial Neural Network are unable to approximate certain functions to any reasonable degree of accuracy. We quantify the complexity of continuous differentiable real valued functions over some interval in order to find the minimum architecture needed to approximate the function over that interval.

Location

Hartman Union Building Courtroom

Start Date

5-2-2019 2:00 PM

End Date

5-2-2019 3:00 PM

This document is currently not available here.

Share

COinS
 
May 2nd, 2:00 PM May 2nd, 3:00 PM

Defining Complexity of Continuous Functions in Relation to Artificial Neural Networks

Hartman Union Building Courtroom

Artificial Neural Networks are general function approximators. Some architectures of an Artificial Neural Network are unable to approximate certain functions to any reasonable degree of accuracy. We quantify the complexity of continuous differentiable real valued functions over some interval in order to find the minimum architecture needed to approximate the function over that interval.