Regularizing Neural Network Training via Identity-wise Discriminative Feature Suppression

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chapman, Avraham, Liu, Lingqiao
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929490183061504
author Chapman, Avraham
Liu, Lingqiao
author_facet Chapman, Avraham
Liu, Lingqiao
contents It is well-known that a deep neural network has a strong fitting capability and can easily achieve a low training error even with randomly assigned class labels. When the number of training samples is small, or the class labels are noisy, networks tend to memorize patterns specific to individual instances to minimize the training error. This leads to the issue of overfitting and poor generalisation performance. This paper explores a remedy by suppressing the network's tendency to rely on instance-specific patterns for empirical error minimisation. The proposed method is based on an adversarial training framework. It suppresses features that can be utilized to identify individual instances among samples within each class. This leads to classifiers only using features that are both discriminative across classes and common within each class. We call our method Adversarial Suppression of Identity Features (ASIF), and demonstrate the usefulness of this technique in boosting generalisation accuracy when faced with small datasets or noisy labels. Our source code is available.
format Preprint
id arxiv_https___arxiv_org_abs_2209_14553
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Regularizing Neural Network Training via Identity-wise Discriminative Feature Suppression
Chapman, Avraham
Liu, Lingqiao
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
It is well-known that a deep neural network has a strong fitting capability and can easily achieve a low training error even with randomly assigned class labels. When the number of training samples is small, or the class labels are noisy, networks tend to memorize patterns specific to individual instances to minimize the training error. This leads to the issue of overfitting and poor generalisation performance. This paper explores a remedy by suppressing the network's tendency to rely on instance-specific patterns for empirical error minimisation. The proposed method is based on an adversarial training framework. It suppresses features that can be utilized to identify individual instances among samples within each class. This leads to classifiers only using features that are both discriminative across classes and common within each class. We call our method Adversarial Suppression of Identity Features (ASIF), and demonstrate the usefulness of this technique in boosting generalisation accuracy when faced with small datasets or noisy labels. Our source code is available.
title Regularizing Neural Network Training via Identity-wise Discriminative Feature Suppression
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2209.14553