papers AI Learner
The Github is limit! Click to go to the new site.

Full-Jacobian Representation of Neural Networks

2019-05-02
Suraj Srinivas, Francois Fleuret

Abstract

Non-linear functions such as neural networks can be locally approximated by affine planes. Recent works make use of input-Jacobians, which describe the normal to these planes. In this paper, we introduce full-Jacobians, which includes this normal along with an additional intercept term called the bias-Jacobians, that together completely describe local planes. For ReLU neural networks, bias-Jacobians correspond to sums of gradients of outputs w.r.t. intermediate layer activations. We first use these full-Jacobians for distillation by aligning gradients of their intermediate representations. Next, we regularize bias-Jacobians alone to improve generalization. Finally, we show that full-Jacobian maps can be viewed as saliency maps. Experimental results show improved distillation on small data-sets, improved generalization for neural network training, and sharper saliency maps.

Abstract (translated by Google)
URL

http://arxiv.org/abs/1905.00780

PDF

http://arxiv.org/pdf/1905.00780


Similar Posts

Comments