Exploring energy and accuracy tradeoff in structure simplification of trained deep neural networks | IEEE Conference Publication | IEEE Xplore