Path-SGD: Path-Normalized Optimization in Deep Neural Networks
This work addresses optimization challenges in deep learning, offering a novel method that improves training efficiency for practitioners.
The authors tackled the problem of optimizing deep neural networks by proposing Path-SGD, a method based on a geometry invariant to weight rescaling, which leads to empirical gains over SGD and AdaGrad.
We revisit the choice of SGD for training deep neural networks by reconsidering the appropriate geometry in which to optimize the weights. We argue for a geometry invariant to rescaling of weights that does not affect the output of the network, and suggest Path-SGD, which is an approximate steepest descent method with respect to a path-wise regularizer related to max-norm regularization. Path-SGD is easy and efficient to implement and leads to empirical gains over SGD and AdaGrad.