Thesis Topic Details

Topic ID:
1013
Title:
Enhancing a backprop package to support sigma-pi connections
Supervisor:
William H. Wilson
Research Area:
Artificial Intelligence
Associated Staff
Assessor:
Nadine Marcus
Topic Details
Status:
Active
Type:
Development
Programs:
CE SE
Group Suitable:
No
Industrial:
No
Pre-requisites:
Having done COMP3411 would be a good idea.
Description:
Error backpropagation is a major neural network learning algorithm. A sigma-pi connection corresponds to what neuroanatomists call an axono-axono-dendritic synapse. Most neural net packages do not support sigma-pi unit learning.
The aim of this project is to add sigma-pi facilities to the tlearn NN package. This will involve learning about the backprop algorithm (regular and sigma-pi versions), getting to know the internals of tlearn, modifying at least the configuration parser and the learning algorithm to handle sigma-pi units, and documenting the enhanced system.
Comments:
--
Past Student Reports
 
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.