Using Cooperative Game Theory to Prune Neural Networks