arXiv research

Study shows adversarial robustness and common perturbation robustness are independent.

In this paper we establish rigorous benchmarks for image classifier robustness. Our first benchmark, ImageNet-C, standardizes and expands the corruption robustness topic, while showing which classifiers are preferable in safety-critical applications. Then we propose a new dataset called ImageNet-P which enables researc…

In this paper we will perturb the scalar curvature of compact Kahler manifolds by incorporating it with higher Chern forms, and then show that the perturbed scalar curvature has many common properties with the unperturbed scalar curvature. In particular the perturbed scalar curvature becomes a moment map, with respect …

Adversarial perturbations fool deepfake detectors with high accuracy.

Minimal token perturbations reveal how Transformer models process information.

We study the relationship between Bar-Natan's perturbation in Khovanov homology and Szabo's geometric spectral sequence, and construct a link invariant that generalizes both into a common theory. We study a few properties of the new invariant, and introduce a family of s-invariants from the new theory in the same spiri…

The paper learns perturbation sets from data to improve robustness in machine learning.

Gaussian copulas are widely used in the industry to correlate two random variables when there is no prior knowledge about the co-dependence between them. The perturbed Gaussian copula approach allows introducing the skew information of both random variables into the co-dependence structure. The analytical expression of…

New regularization techniques improve stability of deep neural networks.

One of the common tasks in unsupervised learning is dimensionality reduction, where the goal is to find meaningful low-dimensional structures hidden in high-dimensional data. Sometimes referred to as manifold learning, this problem is closely related to the problem of localization, which aims at embedding a weighted gr…

We show that if $K$ is a knot in $S^3$ and $Σ$ is a bridge sphere for $K$ with high distance and $2n$ punctures, the number of perturbations of $K$ required to interchange the two balls bounded by $Σ$ via an isotopy is $n$. We also construct a knot with two different bridge spheres with $2n$ and $2n-1$ bridges respecti…

Study reveals class-dependent effects in perturbation-based feature attribution metrics for time series classification.

NoisyMix boosts model robustness to common corruptions.

New technique stabilizes singular values in concatenated matrices.

A key problem in research on adversarial examples is that vulnerability to adversarial examples is usually measured by running attack algorithms. Because the attack algorithms are not optimal, the attack algorithms are prone to overestimating the size of perturbation needed to fool the target model. In other words, the…

CSD learns a common component for domain generalization, outperforming existing methods.

FTPL with Fréchet perturbation achieves near optimal regret bounds for m-set semi-bandit problems.

Proposes a stability evaluation criterion for learning models using distributional perturbations.

New method improves blackbox attack transferability by perturbing feature hierarchy.

PWGF escapes saddle points in nonconvex optimization.

Wider networks improve natural accuracy but worsen perturbation stability, affecting overall robustness.

Paper defends deep learning classifiers against channel-aware adversarial attacks.

Fast algorithm for rescaling vectors with clipping, improving training efficiency.

This paper focuses on spectral filters on graphs, namely filters defined as elementwise multiplication in the frequency domain of a graph. In many graph signal processing settings, it is important to transfer a filter from one graph to another. One example is in graph convolutional neural networks (ConvNets), where the…

State entropy regularization improves robustness in reinforcement learning, especially under structured perturbations.

We regard pre-trained residual networks (ResNets) as nonlinear systems and use linearization, a common method used in the qualitative analysis of nonlinear systems, to understand the behavior of the networks under small perturbations of the input images. We work with ResNet-56 and ResNet-110 trained on the CIFAR-10 dat…

Adversarial training improves robustness against common corruptions.

Adversarial examples are perturbed inputs designed to fool machine learning models. Most recent works on adversarial examples for image classification focus on directly modifying pixels with minor perturbations. A common requirement in all these works is that the malicious perturbations should be small enough (measured…

This study evaluates adversarial attacks and defenses for chest X-ray disease classification.

New method improves robustness of OOD detection models.

Classifiers fail to classify correctly input images that have been purposefully and imperceptibly perturbed to cause misclassification. This susceptability has been shown to be consistent across classifiers, regardless of their type, architecture or parameters. Common defenses against adversarial attacks modify the cla…

This project improves model robustness to affine transformations.

Exact simulation method for market impact estimation under various execution strategies.

New method improves solving combinatorial optimization problems with smoothed policies.

This work develops sampling methods for differential privacy using SHK geometry.

New method speeds up training of deep networks robust to adversarial attacks.

Adversarial examples are malicious inputs crafted to cause a model to misclassify them. Their most common instantiation, "perturbation-based" adversarial examples introduce changes to the input that leave its true label unchanged, yet result in a different model prediction. Conversely, "invariance-based" adversarial ex…

Deep learning models for graphs have achieved strong performance for the task of node classification. Despite their proliferation, currently there is no study of their robustness to adversarial attacks. Yet, in domains where they are likely to be used, e.g. the web, adversaries are common. Can deep learning models for …

New robustness test for kernel goodness-of-fit tests.

Enhances neural network robustness with Mixup and TLAT.

When the planning horizon is long, and the safe asset grows indefinitely, isoelastic portfolios are nearly optimal for investors who are close to isoelastic for high wealth, and not too risk averse for low wealth. We prove this result in a general arbitrage-free, frictionless, semimartingale model. As a consequence, op…

New method learns cell trajectories and network interactions from single-cell data.

PDA improves deep neural networks' robustness against adversarial and common corruptions.

Simple technique turns any adversarial attack into a universal one using few test examples.

Label noise in adversarial training leads to robust overfitting, explained and mitigated.

When the performance of a machine learning model varies over groups defined by sensitive attributes (e.g., gender or ethnicity), the performance disparity can be expressed in terms of the probability distributions of the input and output variables over each group. In this paper, we exploit this fact to reduce the dispa…

Linear classifiers can resist adversarial attacks on Gaussian data.

A policy is said to be robust if it maximizes the reward while considering a bad, or even adversarial, model. In this work we formalize two new criteria of robustness to action uncertainty. Specifically, we consider two scenarios in which the agent attempts to perform an action $a$, and (i) with probability $α$, an alt…