Title :
Pruning within cascade-correlation
Author :
Waugh, Sam ; Adams, Anthony
Author_Institution :
Artificial Neural Network Res. Group, Tasmania Univ., Hobart, Tas., Australia
Abstract :
Cascade-correlation produces networks with all possible feedforward connections. The pruning of redundant connections using a simple saliency measure is examined, along with three methods for stopping the pruning process which may be used with any network type. The methods are shown to be effective and it is further shown that a large number of redundant connections may be located and removed from cascade-correlation networks
Keywords :
feedforward neural nets; learning (artificial intelligence); cascade-correlation networks; feedforward connections; pruning; redundant connections; Artificial neural networks; Benchmark testing; Computer science; DH-HEMTs; Design for experiments; Feedforward systems; Light emitting diodes; Measurement standards; Size measurement; Spirals;
Conference_Titel :
Neural Networks, 1995. Proceedings., IEEE International Conference on
Conference_Location :
Perth, WA
Print_ISBN :
0-7803-2768-3
DOI :
10.1109/ICNN.1995.487325
