Training More Robust Classification Model via Discriminative Loss and Gaussian Noise Injection

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Nguyen, Hai-Vy, Gamboa, Fabrice, Zhang, Sixin, Chhaibi, Reda, Gratton, Serge, Giaccone, Thierry
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915708178268160
author Nguyen, Hai-Vy
Gamboa, Fabrice
Zhang, Sixin
Chhaibi, Reda
Gratton, Serge
Giaccone, Thierry
author_facet Nguyen, Hai-Vy
Gamboa, Fabrice
Zhang, Sixin
Chhaibi, Reda
Gratton, Serge
Giaccone, Thierry
contents Robustness of deep neural networks to input noise remains a critical challenge, as naive noise injection often degrades accuracy on clean (uncorrupted) data. We propose a novel training framework that addresses this trade-off through two complementary objectives. First, we introduce a loss function applied at the penultimate layer that explicitly enforces intra-class compactness and increases the margin to analytically defined decision boundaries. This enhances feature discriminativeness and class separability for clean data. Second, we propose a class-wise feature alignment mechanism that brings noisy data clusters closer to their clean counterparts. Furthermore, we provide a theoretical analysis demonstrating that improving feature stability under additive Gaussian noise implicitly reduces the curvature of the softmax loss landscape in input space, as measured by Hessian eigenvalues.This thus naturally enhances robustness without explicit curvature penalties. Conversely, we also theoretically show that lower curvatures lead to more robust models. We validate the effectiveness of our method on standard benchmarks and our custom dataset. Our approach significantly reinforces model robustness to various perturbations while maintaining high accuracy on clean data, advancing the understanding and practice of noise-robust deep learning.
format Preprint
id arxiv_https___arxiv_org_abs_2405_18499
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Training More Robust Classification Model via Discriminative Loss and Gaussian Noise Injection
Nguyen, Hai-Vy
Gamboa, Fabrice
Zhang, Sixin
Chhaibi, Reda
Gratton, Serge
Giaccone, Thierry
Machine Learning
Robustness of deep neural networks to input noise remains a critical challenge, as naive noise injection often degrades accuracy on clean (uncorrupted) data. We propose a novel training framework that addresses this trade-off through two complementary objectives. First, we introduce a loss function applied at the penultimate layer that explicitly enforces intra-class compactness and increases the margin to analytically defined decision boundaries. This enhances feature discriminativeness and class separability for clean data. Second, we propose a class-wise feature alignment mechanism that brings noisy data clusters closer to their clean counterparts. Furthermore, we provide a theoretical analysis demonstrating that improving feature stability under additive Gaussian noise implicitly reduces the curvature of the softmax loss landscape in input space, as measured by Hessian eigenvalues.This thus naturally enhances robustness without explicit curvature penalties. Conversely, we also theoretically show that lower curvatures lead to more robust models. We validate the effectiveness of our method on standard benchmarks and our custom dataset. Our approach significantly reinforces model robustness to various perturbations while maintaining high accuracy on clean data, advancing the understanding and practice of noise-robust deep learning.
title Training More Robust Classification Model via Discriminative Loss and Gaussian Noise Injection
topic Machine Learning
url https://arxiv.org/abs/2405.18499