Understanding Robust Learning through the Lens of Representation Similarities

06/20/2022
by   Christian Cianfarani, et al.
33

Representation learning, i.e. the generation of representations useful for downstream applications, is a task of fundamental importance that underlies much of the success of deep neural networks (DNNs). Recently, robustness to adversarial examples has emerged as a desirable property for DNNs, spurring the development of robust training methods that account for adversarial examples. In this paper, we aim to understand how the properties of representations learned by robust training differ from those obtained from standard, non-robust training. This is critical to diagnosing numerous salient pitfalls in robust networks, such as, degradation of performance on benign inputs, poor generalization of robustness, and increase in over-fitting. We utilize a powerful set of tools known as representation similarity metrics, across three vision datasets, to obtain layer-wise comparisons between robust and non-robust DNNs with different architectures, training procedures and adversarial constraints. Our experiments highlight hitherto unseen properties of robust representations that we posit underlie the behavioral differences of robust networks. We discover a lack of specialization in robust networks' representations along with a disappearance of `block structure'. We also find overfitting during robust training largely impacts deeper layers. These, along with other findings, suggest ways forward for the design and training of better robust networks.

READ FULL TEXT

page 20

page 25

page 30

page 31

page 32

page 33

page 34

page 35

01/23/2020

Towards Robust DNNs: An Taylor Expansion-Based Method for Generating Powerful Adversarial Examples

Although deep neural networks (DNNs) have achieved successful applicatio...
03/11/2021

Improving Adversarial Robustness via Channel-wise Activation Suppressing

The study of adversarial examples and their activation has attracted sig...
02/02/2022

Finding Biological Plausibility for Adversarially Robust Features via Metameric Tasks

Recent work suggests that representations learned by adversarially robus...
06/06/2019

Understanding Adversarial Behavior of DNNs by Disentangling Non-Robust and Robust Components in Performance Metric

The vulnerability to slight input perturbations is a worrying yet intrig...
05/25/2018

Training verified learners with learned verifiers

This paper proposes a new algorithmic framework,predictor-verifier train...
03/13/2018

Deep k-Nearest Neighbors: Towards Confident, Interpretable and Robust Deep Learning

Deep neural networks (DNNs) enable innovative applications of machine le...
05/30/2023

Pointwise Representational Similarity

With the increasing reliance on deep neural networks, it is important to...

Please sign up or login with your details

Forgot password? Click here to reset