Figure 16. The effect of training with four stimuli upon the information content of the output layer.

It can be seen that before training, the information content according to the single cell (A) and multiple cell (B) information measures is extremely low. After training the network, the single cell information measure shows that a number of cells in the output layer are maximally informative in discriminating between the stimuli. Similarly, the multiple cell information measure also reaches the maximum 2 bits, showing that all four stimuli's transforms have been learnt by the network.