We argue that, when learning a 1-Lipschitz neural network with the dual loss of an optimal transportation problem, the gradient of the model is both the direction of the transportation plan and the direction to the closest adversarial attack.
no code implementations • 24 Mar 2022 • Mohamed Chafik Bakkay, Mathieu Serrurier, Valentin Kivachuk Burda, Florian Dupuy, Naty Citlali Cabrera-Gutierrez, Michael Zamo, Maud-Alix Mader, Olivier Mestre, Guillaume Oller, Jean-Christophe Jouhaud, Laurent Terray
Precipitation nowcasting is of great importance for weather forecast users, for activities ranging from outdoor activities and sports competitions to airport traffic management.
However they remain commonly considered as less accurate, and their properties in learning are still not fully understood.
This loss function has a direct interpretation in terms of adversarial robustness together with certifiable robustness bound.
We propose a model based on the estimation of two-parameter Weibull distribution conditionally to the features.
no code implementations • 13 Dec 2019 • Zied Bouraoui, Antoine Cornuéjols, Thierry Denœux, Sébastien Destercke, Didier Dubois, Romain Guillaume, João Marques-Silva, Jérôme Mengin, Henri Prade, Steven Schockaert, Mathieu Serrurier, Christel Vrain
Some common concerns are identified and discussed such as the types of used representation, the roles of knowledge and data, the lack or the excess of information, or the need for explanations and causal understanding.
In this paper, we investigate how to learn a suitable representation of satellite image time series in an unsupervised manner by leveraging large amounts of unlabeled data.
This work presents a new method to find two-sided predictive intervals for non-parametric least squares regression without the homoscedasticity assumption.