Understanding Why Neural Networks Generalize Well Through GSNR of Parameters

Jinlong Liu, Yunzhi Bai, Guoqing Jiang, Ting Chen, Huayan Wang

Keywords: dnn, generalization, gradient descent, optimization, regression

Wed Session 3 (12:00-14:00 GMT) [Live QA] [Cal]
Wed Session 4 (17:00-19:00 GMT) [Live QA] [Cal]
Wednesday: Optimisation I

Abstract: As deep neural networks (DNNs) achieve tremendous success across many application domains, researchers tried to explore in many aspects on why they generalize well. In this paper, we provide a novel perspective on these issues using the gradient signal to noise ratio (GSNR) of parameters during training process of DNNs. The GSNR of a parameter is simply defined as the ratio between its gradient's squared mean and variance, over the data distribution. Based on several approximations, we establish a quantitative relationship between model parameters' GSNR and the generalization gap. This relationship indicates that larger GSNR during training process leads to better generalization performance. Futher, we show that, different from that of shallow models (e.g. logistic regression, support vector machines), the gradient descent optimization dynamics of DNNs naturally produces large GSNR during training, which is probably the key to DNNs’ remarkable generalization ability.

Similar Papers

Understanding Generalization in Recurrent Neural Networks
Zhuozhuo Tu, Fengxiang He, Dacheng Tao,
The asymptotic spectrum of the Hessian of DNN throughout training
Arthur Jacot, Franck Gabriel, Clement Hongler,