For context, if true (the paper has not been peer reviewed yet), this confirms what has already been suspected and empirically evaluated about deep learning: th

For context, if true (the paper has not been peer reviewed yet), this confirms what has already been suspected and empirically evaluated about deep learning: that the non-convexity of neural networks is not an issue.