A pruning technique maximizing generalization
Pietro Burrascano · 2005
The paper proposes an estimator of the relevance of each weight connection on generalization which relies on a probabilistic description of the training phase: based on the above estimator, a pruning technique is derived.