Auratus A Java Based Neural Network Suite

Date of Award

2003

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

Mullins, David

Keywords

Java, Neural Network(s), Framework

Area of Concentration

Software Systems

Abstract

As the computational power of modern computers continues to increase, software developers have the ability to research more complex topologies and problem domains for artificial neural networks. Proper software design is a crucial step in allowing developers to reduce development time and expenses, ideally providing a set of architecture suggestions and pre-built components that allow developers to create real-world applications with diminished effort. This thesis focuses on the and evaluation of the Auratus software framework: a java based framework for neural network development. The Auratus framework contains a set of node and edge components, as well as a set of derived components representing the network, algorithm, and other various high-level constructs, commonly used in the development of neural networks. Auratus succeeds in providing a set of reusable and extendible controls, allowing for faster development of neural network applications. However, developers may find it difficult to adapt to Auratus' design, especially if they are use to classical design methodologies. Although construction of algorithms and network topologies is highly intuitive with the Auratus framework, developers may find the additional resource requirements Auratus requires too high.

Rights

This bibliographic record is available under the Creative Commons CC0 public domain dedication. The New College of Florida, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.

This document is currently not available here.

Share

COinS