Abstract
In this paper, we propose a simple multilayer cluster neural network with five independent subnetworks for off-line recognition of totally unconstrained handwritten numerals. We also show that the use of genetic algorithms for avoiding the problem of finding local minima in training the multilayer cluster neural network with gradient descent technique reduces error rates.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the 12th IAPR International Conference on Pattern Recognition - Conference B |
| Subtitle of host publication | Pattern Recognition and Neural Networks, ICPR 1994 |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 507-509 |
| Number of pages | 3 |
| ISBN (Electronic) | 0818662700 |
| Publication status | Published - 1994 |
| Externally published | Yes |
| Event | 12th IAPR International Conference on Pattern Recognition - Conference B: Pattern Recognition and Neural Networks, ICPR 1994 - Jerusalem, Israel Duration: 1994 Oct 9 → 1994 Oct 13 |
Publication series
| Name | Proceedings - International Conference on Pattern Recognition |
|---|---|
| Volume | 2 |
| ISSN (Print) | 1051-4651 |
Conference
| Conference | 12th IAPR International Conference on Pattern Recognition - Conference B: Pattern Recognition and Neural Networks, ICPR 1994 |
|---|---|
| Country/Territory | Israel |
| City | Jerusalem |
| Period | 94/10/9 → 94/10/13 |
Bibliographical note
Funding Information:This research was supported by the 1992 Directed Basic Research Fund of Korea Science and Engineering Foundation.
Funding Information:
bined a genetic algorithm with the multilayer cluster neural network to avoid the problem of finding local minima in training with a gradient descent technique. Consequently, the use of a genetic algorithm reduced error rates. In this paper, we used a simple multilayer cluster neural network which has 10 output units: one per class. However, considering multiple models for the class which has wide variations, it is expected that the performance of proposed scheme will be improved. Further investigation should be made, however, to design a locally constrained cluster network architec-ture which has good generalization and involves mul-tiple models and to develop a technique in which segmentation and recognition are integrated for the recognition of unconstrained handwritten, connected numerals. Acknowledgments This research was supported by the 1992 Directed Basic Research Fund of Korea Science and Engineer-ing Foundation.
Publisher Copyright:
© 1994 Institute of Electrical and Electronics Engineers Inc.. All rights reserved.
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition