| Author: | Ma, Wing-hong |
| Title: | A study of conjugate gradient recurrent network on time series problems |
| Degree: | M.Sc. |
| Year: | 1997 |
| Subject: | Time-series analysis Neural networks (Computer science) Real-time control Hong Kong Polytechnic University -- Dissertations |
| Department: | Multi-disciplinary Studies |
| Pages: | vi, 36 leaves : ill. ; 30 cm |
| Language: | English |
| Abstract: | Recurrent networks, which include feedback loops, are capable of processing temporal patterns and accepting sequences as inputs and producing them as outputs. Recurrent networks can be trained with error backpropagation and real-time recurrent learning. But these training algorithms require many training cycles for the networks to converge. Learning algorithm based on second order method such as conjugate gradient had been successfully applied to feedforward and recurrent networks and the number of epochs needed for training was significantly reduced. However, the conjugate gradient method can only be run in batch mode; this mades it unsuitable for time series problems. In this dissertation, a modified conjugate gradient recurrent learning algorithm, which is able to run in sequence mode, is proposed to tackle the time series problems. The Henon series was used to test the performance of the network and the results were compared with the modified steepest gradient recurrent learning algorithm, a first order recurrent learning algorithm. Finally, the results were also compared with varies form of RTRL. Results shows that modified CGRL network can learn the Henon series successfully. It also had performance advantages over modified SGRL and varies form of RTRL in terms number of epochs required for convergence. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| b13986429.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 1.69 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/2250