Salvato in:
Dettagli Bibliografici
Autori principali: Wu, Shangxi, Lu, Dongyuan, Zhao, Xian, Chen, Lizhang, Sang, Jitao
Natura: Preprint
Pubblicazione: 2020
Soggetti:
Accesso online:https://arxiv.org/abs/2011.09789
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909224784625664
author Wu, Shangxi
Lu, Dongyuan
Zhao, Xian
Chen, Lizhang
Sang, Jitao
author_facet Wu, Shangxi
Lu, Dongyuan
Zhao, Xian
Chen, Lizhang
Sang, Jitao
contents Deep learning models suffer from the problem of semantic discontinuity: small perturbations in the input space tend to cause semantic-level interference to the model output. We argue that the semantic discontinuity results from these inappropriate training targets and contributes to notorious issues such as adversarial robustness, interpretability, etc. We first conduct data analysis to provide evidence of semantic discontinuity in existing deep learning models, and then design a simple semantic continuity constraint which theoretically enables models to obtain smooth gradients and learn semantic-oriented features. Qualitative and quantitative experiments prove that semantically continuous models successfully reduce the use of non-semantic information, which further contributes to the improvement in adversarial robustness, interpretability, model transfer, and machine bias.
format Preprint
id arxiv_https___arxiv_org_abs_2011_09789
institution arXiv
publishDate 2020
record_format arxiv
spellingShingle An Experimental Study of Semantic Continuity for Deep Learning Models
Wu, Shangxi
Lu, Dongyuan
Zhao, Xian
Chen, Lizhang
Sang, Jitao
Machine Learning
Computer Vision and Pattern Recognition
Deep learning models suffer from the problem of semantic discontinuity: small perturbations in the input space tend to cause semantic-level interference to the model output. We argue that the semantic discontinuity results from these inappropriate training targets and contributes to notorious issues such as adversarial robustness, interpretability, etc. We first conduct data analysis to provide evidence of semantic discontinuity in existing deep learning models, and then design a simple semantic continuity constraint which theoretically enables models to obtain smooth gradients and learn semantic-oriented features. Qualitative and quantitative experiments prove that semantically continuous models successfully reduce the use of non-semantic information, which further contributes to the improvement in adversarial robustness, interpretability, model transfer, and machine bias.
title An Experimental Study of Semantic Continuity for Deep Learning Models
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2011.09789