no code implementations • 9 Oct 2023 • Tim De Ryck, Florent Bonnet, Siddhartha Mishra, Emmanuel de Bézenac
In this paper, we investigate the behavior of gradient descent algorithms in physics-informed machine learning methods like PINNs, which minimize residuals connected to partial differential equations (PDEs).
1 code implementation • NeurIPS 2023 • Jean-Yves Franceschi, Mike Gartrell, Ludovic Dos Santos, Thibaut Issenhuth, Emmanuel de Bézenac, Mickaël Chen, Alain Rakotomamonjy
Particle-based deep generative models, such as gradient flows and score-based diffusion models, have recently gained traction thanks to their striking performance.
1 code implementation • NeurIPS 2023 • Bogdan Raonić, Roberto Molinaro, Tim De Ryck, Tobias Rohner, Francesca Bartolucci, Rima Alaifari, Siddhartha Mishra, Emmanuel de Bézenac
Although very successfully used in conventional machine learning, convolution based neural network architectures -- believed to be inconsistent in function space -- have been largely ignored in the context of learning solution operators of PDEs.
no code implementations • 3 Oct 2022 • Skander Karkar, Ibrahim Ayed, Emmanuel de Bézenac, Patrick Gallinari
End-to-end backpropagation has a few shortcomings: it requires loading the entire model during training, which can be impossible in constrained settings, and suffers from three locking problems (forward locking, update locking and backward locking), which prohibit training the layers in parallel.
1 code implementation • 10 Jun 2021 • Jean-Yves Franceschi, Emmanuel de Bézenac, Ibrahim Ayed, Mickaël Chen, Sylvain Lamprier, Patrick Gallinari
We propose a novel theoretical framework of analysis for Generative Adversarial Networks (GANs).
1 code implementation • NeurIPS 2021 • Yuan Yin, Ibrahim Ayed, Emmanuel de Bézenac, Nicolas Baskiotis, Patrick Gallinari
Both are sub-optimal: the former disregards the discrepancies between environments leading to biased solutions, while the latter does not exploit their potential commonalities and is prone to scarcity problems.
no code implementations • NeurIPS 2020 • Richard Kurle, Syama Sundar Rangapuram, Emmanuel de Bézenac, Stephan Günnemann, Jan Gasthaus
We propose a Monte Carlo objective that leverages the conditional linearity by computing the corresponding conditional expectations in closed-form and a suitable proposal distribution that is factorised similarly to the optimal proposal distribution.
no code implementations • NeurIPS 2020 • Emmanuel de Bézenac, Syama Sundar Rangapuram, Konstantinos Benidis, Michael Bohlke-Schneider, Richard Kurle, Lorenzo Stella, Hilaf Hasson, Patrick Gallinari, Tim Januschowski
This paper tackles the modelling of large, complex and multivariate time series panels in a probabilistic setting.
2 code implementations • ICLR 2021 • Yuan Yin, Vincent Le Guen, Jérémie Dona, Emmanuel de Bézenac, Ibrahim Ayed, Nicolas Thome, Patrick Gallinari
In this work, we introduce the APHYNITY framework, a principled approach for augmenting incomplete physical dynamics described by differential equations with deep data-driven models.
1 code implementation • 17 Sep 2020 • Skander Karkar, Ibrahim Ayed, Emmanuel de Bézenac, Patrick Gallinari
From this observation, we reformulate the learning problem as follows: finding neural networks which solve the task while transporting the data as efficiently as possible.
no code implementations • 25 Sep 2019 • Yuan Yin, Arthur Pajot, Emmanuel de Bézenac, Patrick Gallinari
We tackle the problem of inpainting occluded area in spatiotemporal sequences, such as cloud occluded satellite observations, in an unsupervised manner.
no code implementations • 4 Jun 2019 • Emmanuel de Bézenac, Ibrahim Ayed, Patrick Gallinari
Domain Translation is the problem of finding a meaningful correspondence between two domains.
no code implementations • ICLR 2019 • Ibrahim Ayed, Emmanuel de Bézenac, Arthur Pajot, Patrick Gallinari
Spatio-Temporal processes bear a central importance in many applied scientific fields.
no code implementations • 26 Feb 2019 • Ibrahim Ayed, Emmanuel de Bézenac, Arthur Pajot, Julien Brajard, Patrick Gallinari
We consider the problem of forecasting complex, nonlinear space-time processes when observations provide only partial information of on the system's state.