ИСТИНА |
Войти в систему Регистрация |
|
ИСТИНА ИНХС РАН |
||
We apply different modifications of Fahlman and Lebier’s cascade correlation algorithm to a real-world application in order to fulfil a comparison of these approaches in terms of generalisation ability and time elapsed for training. Investigation shows that the algorithm should be varied in view of different statements of training and retraining problems for optimal performance.