Аннотация
EPNet is an evolutionary system for automatic design of artificial
neural networks (ANNs) 1, 2, 3. Unlike most previous methods on
evolving ANNs, EPNet puts its emphasis on evolving ANN'S behaviours
rather than circuitry. The parsimony of evolved ANNs is encouraged
by the sequential application of architectural mutations. In this
paper, EP Net is applied to a couple of chaotic time-series prediction
problems (i.e., the Mackey-Glass differential equation and the logistic
map). The experimental results show that EPNet can produce very compact
ANNs with good prediction ability in comparison with other algorithms.
Пользователи данного ресурса
Пожалуйста,
войдите в систему, чтобы принять участие в дискуссии (добавить собственные рецензию, или комментарий)