Publications de Seyed Mohsen Moosavi Dezfooli

Certified Human Trajectory Prediction

Alexandre Massoud Alahi, Seyed Mohsen Moosavi Dezfooli, Saeed Saadatnejad, Mohammadhossein Bahari

Predicting human trajectories is essential for the safe operation of autonomous vehicles, yet current data-driven models often lack robustness in case of noisy inputs such as adversarial examples or imperfect observations. Although some trajectory predicti ...

2025

Sparse Attacks for Manipulating Explanations in Deep Neural Network Models

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Michail Vlachos, Ahmad Ajalloeian

We investigate methods for manipulating classifier explanations while keeping the predictions unchanged. Our focus is on using a sparse attack, which seeks to alter only a minimal number of input features. We present an efficient and novel algorithm for co ...

Ieee Computer Soc2023

On Smoothed Explanations: Quality and Robustness

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Michail Vlachos, Ahmad Ajalloeian

Explanation methods highlight the importance of the input features in taking a predictive decision, and represent a solution to increase the transparency and trustworthiness in machine learning and deep neural networks (DNNs). However, explanation methods ...

Assoc Computing Machinery2022

Are socially-aware trajectory prediction models really socially-aware?

Alexandre Massoud Alahi, Seyed Mohsen Moosavi Dezfooli, Saeed Saadatnejad, Mohammadhossein Bahari, Pedram Khorsandi

Our transportation field has recently witnessed an arms race of neural network-based trajectory predictors. While these predictors are at the core of many applications such as autonomous navigation or pedestrian flow simulations, their adversarial robustne ...

2022

Vehicle trajectory prediction works, but not everywhere

Alexandre Massoud Alahi, Seyed Mohsen Moosavi Dezfooli, Saeed Saadatnejad, Ahmad Rahimi, Mohammadhossein Bahari

Vehicle trajectory prediction is nowadays a fundamental pillar of self-driving cars. Both the industry and research communities have acknowledged the need for such a pillar by providing public benchmarks. While state-of-the-art methods are impressive, i.e. ...

IEEE2022

PRIME: A Few Primitives Can Boost Robustness to Common Corruptions

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Guillermo Ortiz Jimenez, Apostolos Modas

Despite their impressive performance on image classification tasks, deep networks have a hard time generalizing to unforeseen corruptions of their data. To fix this vulnerability, prior works have built complex data augmentation strategies, combining multi ...

SPRINGER INTERNATIONAL PUBLISHING AG2022

What can linearized neural networks actually say about generalization?

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Guillermo Ortiz Jimenez

For certain infinitely-wide neural networks, the neural tangent kernel (NTK) theory fully characterizes generalization, but for the networks used in practice, the empirical NTK only provides a rough first-order approximation. Still, a growing body of work ...

2021

Neural Anisotropy Directions

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Guillermo Ortiz Jimenez, Apostolos Modas

In this work, we analyze the role of the network architecture in shaping the inductive bias of deep classifiers. To that end, we start by focusing on a very simple problem, i.e., classifying a class of linearly separable distributions, and show that, depen ...

2020

Optimism in the Face of Adversity: Understanding and Improving Deep Learning Through Adversarial Robustness

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Guillermo Ortiz Jimenez, Apostolos Modas

Driven by massive amounts of data and important advances in computational resources, new deep learning systems have achieved outstanding results in a large spectrum of applications. Nevertheless, our current theoretical understanding on the mathematical fo ...

2020

Hold me tight! Influence of discriminative features on deep network boundaries

Pascal Frossard, Seyed Mohsen Moosavi Dezfooli, Guillermo Ortiz Jimenez, Apostolos Modas

Important insights towards the explainability of neural networks reside in the characteristics of their decision boundaries. In this work, we borrow tools from the field of adversarial robustness, and propose a new perspective that relates dataset features ...

2020