Training neocognitron to recognize handwritten digits in the real world

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. recognise 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.