Salvato in:
Dettagli Bibliografici
Autori principali: Zhou, Linjiang, Ma, Chao, Wang, Zepeng, Wu, Libing, Shi, Xiaochuan
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.07711
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917033019441152
author Zhou, Linjiang
Ma, Chao
Wang, Zepeng
Wu, Libing
Shi, Xiaochuan
author_facet Zhou, Linjiang
Ma, Chao
Wang, Zepeng
Wu, Libing
Shi, Xiaochuan
contents Gradient Smoothing is an efficient approach to reducing noise in gradient-based model explanation method. SmoothGrad adds Gaussian noise to mitigate much of these noise. However, the crucial hyper-parameter in this method, the variance $σ$ of Gaussian noise, is set manually or with heuristic approach. However, it results in the smoothed gradients still containing a certain amount of noise. In this paper, we aim to interpret SmoothGrad as a corollary of convolution, thereby re-understanding the gradient noise and the role of $σ$ from the perspective of confidence level. Furthermore, we propose an adaptive gradient smoothing method, AdaptGrad, based on these insights. Through comprehensive experiments, both qualitative and quantitative results demonstrate that AdaptGrad could effectively reduce almost all the noise in vanilla gradients compared with baselines methods. AdaptGrad is simple and universal, making it applicable for enhancing gradient-based interpretability methods for better visualization.
format Preprint
id arxiv_https___arxiv_org_abs_2410_07711
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle AdaptGrad: Adaptive Sampling to Reduce Noise
Zhou, Linjiang
Ma, Chao
Wang, Zepeng
Wu, Libing
Shi, Xiaochuan
Machine Learning
Gradient Smoothing is an efficient approach to reducing noise in gradient-based model explanation method. SmoothGrad adds Gaussian noise to mitigate much of these noise. However, the crucial hyper-parameter in this method, the variance $σ$ of Gaussian noise, is set manually or with heuristic approach. However, it results in the smoothed gradients still containing a certain amount of noise. In this paper, we aim to interpret SmoothGrad as a corollary of convolution, thereby re-understanding the gradient noise and the role of $σ$ from the perspective of confidence level. Furthermore, we propose an adaptive gradient smoothing method, AdaptGrad, based on these insights. Through comprehensive experiments, both qualitative and quantitative results demonstrate that AdaptGrad could effectively reduce almost all the noise in vanilla gradients compared with baselines methods. AdaptGrad is simple and universal, making it applicable for enhancing gradient-based interpretability methods for better visualization.
title AdaptGrad: Adaptive Sampling to Reduce Noise
topic Machine Learning
url https://arxiv.org/abs/2410.07711