|
Once trained, the network will be tested on all of the sentences of the language. How close the network's output is to the intended output will be measured by adding the square of the difference between the intended and actual output at each output node. The final network error will be the sum of the error for each sentence. The fitness of the trained network will be the inverse of the final error. The fitness of the original genotype will be equal to the fitness of the trained network corrected to account for additional training data and number of connections, as described in the previous section.
|
Back to the previous subject
Back to the Table of Content
To the next subject