Search Results for author: Alexandre Araujo

Found 21 papers, 12 papers with code

PAL: Proxy-Guided Black-Box Attack on Large Language Models

1 code implementation15 Feb 2024 Chawin Sitawarin, Norman Mu, David Wagner, Alexandre Araujo

In this work, we introduce the Proxy-Guided Attack on LLMs (PAL), the first optimization-based attack on LLMs in a black-box query-only setting.

Novel Quadratic Constraints for Extending LipSDP beyond Slope-Restricted Activations

no code implementations25 Jan 2024 Patricia Pauli, Aaron Havens, Alexandre Araujo, Siddharth Garg, Farshad Khorrami, Frank Allgöwer, Bin Hu

However, a direct application of LipSDP to the resultant residual ReLU networks is conservative and even fails in recovering the well-known fact that the MaxMin activation is 1-Lipschitz.

Towards Real-World Focus Stacking with Deep Learning

1 code implementation29 Nov 2023 Alexandre Araujo, Jean Ponce, Julien Mairal

Focus stacking is widely used in micro, macro, and landscape photography to reconstruct all-in-focus images from multiple frames obtained with focus bracketing, that is, with shallow depth of field and different focus planes.

LipSim: A Provably Robust Perceptual Similarity Metric

1 code implementation27 Oct 2023 Sara Ghazanfari, Alexandre Araujo, Prashanth Krishnamurthy, Farshad Khorrami, Siddharth Garg

On the other hand, as perceptual metrics rely on neural networks, there is a growing concern regarding their resilience, given the established vulnerability of neural networks to adversarial attacks.

Image Retrieval Retrieval

Certification of Deep Learning Models for Medical Image Segmentation

1 code implementation5 Oct 2023 Othmane Laousy, Alexandre Araujo, Guillaume Chassagnon, Nikos Paragios, Marie-Pierre Revel, Maria Vakalopoulou

In this paper, we present for the first time a certified segmentation baseline for medical imaging based on randomized smoothing and diffusion models.

Denoising Image Segmentation +3

R-LPIPS: An Adversarially Robust Perceptual Similarity Metric

1 code implementation27 Jul 2023 Sara Ghazanfari, Siddharth Garg, Prashanth Krishnamurthy, Farshad Khorrami, Alexandre Araujo

In this paper, we propose the Robust Learned Perceptual Image Patch Similarity (R-LPIPS) metric, a new metric that leverages adversarially trained deep features.

Towards Better Certified Segmentation via Diffusion Models

1 code implementation16 Jun 2023 Othmane Laousy, Alexandre Araujo, Guillaume Chassagnon, Marie-Pierre Revel, Siddharth Garg, Farshad Khorrami, Maria Vakalopoulou

The robustness of image segmentation has been an important research topic in the past few years as segmentation models have reached production-level accuracy.

Autonomous Driving Image Segmentation +2

Efficient Bound of Lipschitz Constant for Convolutional Layers by Gram Iteration

1 code implementation25 May 2023 Blaise Delattre, Quentin Barthélemy, Alexandre Araujo, Alexandre Allauzen

Since the control of the Lipschitz constant has a great impact on the training stability, generalization, and robustness of neural networks, the estimation of this value is nowadays a real scientific challenge.

Diffusion-Based Adversarial Sample Generation for Improved Stealthiness and Controllability

1 code implementation NeurIPS 2023 Haotian Xue, Alexandre Araujo, Bin Hu, Yongxin Chen

Neural networks are known to be susceptible to adversarial samples: small variations of natural examples crafted to deliberately mislead the models.

Towards Evading the Limits of Randomized Smoothing: A Theoretical Analysis

no code implementations3 Jun 2022 Raphael Ettedgui, Alexandre Araujo, Rafael Pinot, Yann Chevaleyre, Jamal Atif

We first show that these certificates use too little information about the classifier, and are in particular blind to the local curvature of the decision boundary.

A Dynamical System Perspective for Lipschitz Neural Networks

no code implementations25 Oct 2021 Laurent Meunier, Blaise Delattre, Alexandre Araujo, Alexandre Allauzen

The Lipschitz constant of neural networks has been established as a key quantity to enforce the robustness to adversarial examples.

Building Compact and Robust Deep Neural Networks with Toeplitz Matrices

no code implementations2 Sep 2021 Alexandre Araujo

This thesis focuses on the problem of training neural networks which are not only accurate but also compact, easy to train, reliable and robust to adversarial examples.

Advocating for Multiple Defense Strategies against Adversarial Examples

no code implementations4 Dec 2020 Alexandre Araujo, Laurent Meunier, Rafael Pinot, Benjamin Negrevergne

It has been empirically observed that defense mechanisms designed to protect neural networks against $\ell_\infty$ adversarial examples offer poor performance against $\ell_2$ adversarial examples and vice versa.

On Lipschitz Regularization of Convolutional Layers using Toeplitz Matrix Theory

2 code implementations15 Jun 2020 Alexandre Araujo, Benjamin Negrevergne, Yann Chevaleyre, Jamal Atif

This paper tackles the problem of Lipschitz regularization of Convolutional Neural Networks.

The Expressive Power of Deep Neural Networks with Circulant Matrices

no code implementations ICLR 2019 Alexandre Araujo, Benjamin Negrevergne, Yann Chevaleyre, Jamal Atif

Recent results from linear algebra stating that any matrix can be decomposed into products of diagonal and circulant matrices has lead to the design of compact deep neural network architectures that perform well in practice.

General Classification Video Classification

Robust Neural Networks using Randomized Adversarial Training

no code implementations25 Mar 2019 Alexandre Araujo, Laurent Meunier, Rafael Pinot, Benjamin Negrevergne

This paper tackles the problem of defending a neural network against adversarial attacks crafted with different norms (in particular $\ell_\infty$ and $\ell_2$ bounded adversarial examples).

Understanding and Training Deep Diagonal Circulant Neural Networks

no code implementations29 Jan 2019 Alexandre Araujo, Benjamin Negrevergne, Yann Chevaleyre, Jamal Atif

In this paper, we study deep diagonal circulant neural networks, that is deep neural networks in which weight matrices are the product of diagonal and circulant ones.

Video Classification

Cannot find the paper you are looking for? You can Submit a new open access paper.