 |
|
| Fig. 2.1 - Patterns 0 and 1 used for learning | |
On this simple example we will demonstrate abilities of the neocognitron at recognition of presented patterns.
The black-box in this example contains neocognitron network which can distinguish between two different types of patterns (between digit zero and digit one). For the learning of the network we have used patterns shown in figure 2.1.
|
|
| Fig. 2.1 - Patterns 0 and 1 used for learning | |
By the learning of the neocognitron to distinguish between these two types of patterns we have created two different categories in the network. In the future the network will respond on every presented pattern with a pair of values. Each of these values is a measure of belonging of presented pattern into one of two created categories.
Click on one of the prepared patterns. The network processes it and assigns it to one of the categories. Notice, that the network assigns patterns which have not been presented during learning to the correct category, too. These patterns were produced by distortion of patterns used for learning, shown in figure 2.1.