Fetching the paper…

BPGrad: Towards Global Optimality in Deep Learning via Branch and Pruning · Around