The neocognitron is a hierarchical multilayered neural network proposed by Professor Kunihiko Fukushima for handwritten character recognition.
At present there are many different versions of the neocognitron. Two original basic versions proposed by Professor Fukushima differ in used learning principle mainly :
The first version of the neocognitron was based on the learning without a teacher. This version is often called self-organized neocognitron. In this tutorial, however, we will focus on the version of the neocognitron which is based on the learning with a teacher. We believe that this version is more suitable for presentation of the basic principle of the neocognitron.
The main advantage of neocognitron is its ability to recognize correctly not only learned patterns but also patterns which are produced from them by using of partial shift, rotation or another type of distortion.
We will demonstrate abilities of the neocognitron on the following simple example.