15. Example - Learning

In this example we will demonstrate learning of the simple version of the neocognitron network. We want that our network will extract features shown in figure 15.1.

Fig. 15.1 - Hierarchy of extracted features

We prepare the corresponding training pattern set (it is shown quite in the left in example) and we start learning by Start button.

In our demonstration b-weights are not shown because the way of their modification is not so important for us now. Remember however that both a-weights and b-weights are adjusted during learning. Notice that together with seed cell selecting their connection areas and receptive field are selected as well.