The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9746

Title:Investigating Backpropagation Through Time
Date:Sep 1997
Abstract:A lot of industrial and financial applications depend on time. Time-dependent problems deal with every kind of prediction such as stock market prediction. The prediction for the future is based on the progress of the past. We examine what has happened in past years and we decide about the next years. These days, there is a great interest in developing learning algorithms in order to model phenomena dependent on time. In this project, we use the time-dependent backpropagation algorithm to learn termporal trajectories. the mathematical description of the algorithm has been explicated in details and its applications to the XOR problem and in producing circular trajectories are presented.

[Search These Pages] [DAI Home Page] [Comment]