Publications associées à Rétropropagation du gradient

Leveraging Continuous Time to Understand Momentum When Training Diagonal Linear Networks

Nicolas Henri Bernard Flammarion, Hristo Georgiev Papazov, Scott William Pesme

In this work, we investigate the effect of momentum on the optimisation trajectory of gradient descent. We leverage a continuous-time approach in the analysis of momentum gradient descent with step size

\gamma

and momentum parameter

\beta

that allows u ...

2024

Deep Learning Theory Through the Lens of Diagonal Linear Networks

Scott William Pesme

In this PhD manuscript, we explore optimisation phenomena which occur in complex neural networks through the lens of

2

-layer diagonal linear networks. This rudimentary architecture, which consists of a two layer feedforward linear network with a diagonal ...

EPFL2024

Safe Deep Neural Networks

Kyle Michael Matoba

The capabilities of deep learning systems have advanced much faster than our ability to understand them. Whilst the gains from deep neural networks (DNNs) are significant, they are accompanied by a growing risk and gravity of a bad outcome. This is troubli ...

EPFL2024

Efficient local linearity regularization to overcome catastrophic overfitting

Volkan Cevher, Grigorios Chrysos, Fanghui Liu, Elias Abad Rocamora

Catastrophic overfitting (CO) in single-step adversarial training (AT) results in abrupt drops in the adversarial test accuracy (even down to 0%). For models trained with multi-step AT, it has been observed that the loss function behaves locally linearly w ...

2024

Layer-Wise Learning Framework for Efficient DNN Deployment in Biomedical Wearable Systems

David Atienza Alonso, Amir Aminifar, Tomas Teijeiro Campo, Alireza Amirshahi, Saleh Baghersalimi

The development of low-power wearable systems requires specialized techniques to accommodate their unique requirements and constraints. While significant advancements have been made in the inference phase of artificial intelligence, the training phase rema ...

2023

A Neural-Network-Based Convex Regularizer for Inverse Problems

Michaël Unser, Pakshal Narendra Bohra, Alexis Marie Frederic Goujon, Sebastian Jonas Neumayer, Stanislas Ducotterd

The emergence of deep-learning-based methods to solve image-reconstruction problems has enabled a significant increase in quality. Unfortunately, these new methods often lack reliability and explainability, and there is a growing interest to address these ...

2023

Artificial Neural Network Training on an Optical Processor via Direct Feedback Alignment

Florent Gérard Krzakala, Julien Marcel Daniel Emmanuel Launay

Artificial Neural Networks (ANN) are habitually trained via the back-propagation (BP) algorithm. This approach has been extremely successful: Current models like GPT-3 have O(10 11 ) parameters, are trained on O(10 11 ) words and produce awe-inspiring resu ...

IEEE2023

Inverse Reinforcement Learning of Pedestrian-Robot Coordination

Aude Billard, David Julian Gonon

We apply inverse reinforcement learning (IRL) with a novel cost feature to the problem of robot navigation in human crowds. Consistent with prior empirical work on pedestrian behavior, the feature anticipates collisions between agents. We efficiently learn ...

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC2023

Exploring neural networks and their potential for species distribution modeling

Devis Tuia, Benjamin Alexander Kellenberger, Nina Marion Aurélia Van Tiel, Robin Adrien Zbinden, Lloyd Haydn Hughes

Species distribution models (SDMs) relate species occurrence data with environmental variables and are used to understand and predict species distributions across landscapes. While some machine learning models have been adopted by the SDM community, recent ...

2023

Machine Learning Methods for Robust Uncertainty Quantification and Controller Approximation

Emilio Maddalena

This thesis is situated at the crossroads between machine learning and control engineering. Our contributions are both theoretical, through proposing a new uncertainty quantification methodology in a kernelized context; and experimental, through investigat ...

EPFL2023