Model Optimization

RSO: A Gradient Free Sampling Based Approach For Training Deep Neural Networks