Training neocognitron to recognize handwritten digits in the real world
- 22 November 2002
- proceedings article
- Published by Institute of Electrical and Electronics Engineers (IEEE)
Abstract
Using a large scale real-world database ETL-1, we show that the neocognitron trained by unsupervised learning with a winner-take-all process can recognize handwritten digits with a recognition rate higher than 97%. We use the technique of dual thresholds for feature-extracting S-cells, and higher threshold values are used in the learning than in the recognition phase. We discuss how the threshold values affect the recognition rate. The learning method for the cells of the highest stage of the network has been modified from the conventional one, in order to reconcile the unsupervised learning with the use of information of the category names of the training patterns.Keywords
This publication has 7 references indexed in Scilit:
- Neocognitron: A new algorithm for pattern recognition tolerant of deformations and shifts in positionPublished by Elsevier ,2003
- Neocognitron: A hierarchical neural network capable of visual pattern recognitionPublished by Elsevier ,2003
- Improved neocognitron with bend-detecting cellsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2003
- Training neocognitron to recognize handwritten digits in the real worldPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2002
- Use of different thresholds in learning and recognitionNeurocomputing, 1996
- How to adjust the thresholds of feature-extracting cells in the neocognitronSystems and Computers in Japan, 1994
- Analysis of the process of visual pattern recognition by the neocognitronNeural Networks, 1989