paper-with-me

Papers

Understanding Why Neural Networks Generalize Well Through GSNR of Parameters

2020-01-21 · ICLR 2020 1 · Jinlong Liu, Guoqing Jiang, Yunzhi Bai, Ting Chen, Huayan Wang

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 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. Moreover, 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.

📄 PDF Abstract BibTeX arXiv:2001.07384

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Unleashing the Power of Gradient Signal-to-Noise Ratio for Zero-Shot NAS

2023-01-01 · ICCV 2023 1 · Zihao Sun, Yu Sun, Longxing Yang, Shun Lu 외

Neural Architecture Search (NAS) aims to automatically find optimal neural network architectures in an efficient way. Zero-Shot NAS is a promising technique that leverages proxies to predict the accuracy of candidate…

Neural Architecture Search

Domain Generalization Guided by Gradient Signal to Noise Ratio of Parameters

2023-10-11 · ICCV 2023 1 · Mateusz Michalkiewicz, Masoud Faraki, Xiang Yu, Manmohan Chandraker 외

Overfitting to the source domain is a common issue in gradient-based training of deep neural networks. To compensate for the over-parameterized models, numerous regularization techniques have been introduced such as thos…

Domain GeneralizationFace Anti-SpoofingMeta-Learning

Fast WDM provisioning with minimal probing: the first field experiments for DC exchanges

2023-09-14 · Hideki Nishizawa, Toru Mano, Thomas Ferreira de Lima, Yue-Kai Huang 외

We propose an approach to estimate the end-to-end GSNR accurately in a short time when a data center interconnect (DCI) network operator receives a service request from users, not by measuring the GSNR at the operational…

GSNR: Graph Smooth Null-Space Representation for Inverse Problems

2026-02-23 · Romario Gualdrón-Hurtado, Roman Jacome, Rafael S. Suarez, Henry Arguello arxiv

Inverse problems in imaging are ill-posed, leading to infinitely many solutions consistent with the measurements due to the non-trivial null-space of the sensing matrix. Common image priors promote solutions on the gener…

Image Super-ResolutionImage Deblurring

Launch Power Optimization for Dynamic Elastic Optical Networks over C+L Bands

2023-08-25 · Farhad Arpanaei, Mahdi Ranjbar Zefreh, José Alberto Hernández, Behnam Shariati 외

We propose an algorithm for calculating the optimum launch power over the entire C+L bands by maximizing the cumulative link GSNR of a channel plan built upon multiple modulation formats, with application to dynamic EONs…