Simultaneous Weight and Architecture Optimization for Neural Networks