-
Balázs Kégl (Paris-Saclay Center for Data Science)30/03/2015 14:00
-
M. Nicolas Gac (L2S)30/03/2015 14:15Les algorithmes itératifs utilisés lors de la résolution de problèmes inverses portant sur des gros volume de données requiert une accélération significative pour être utilisés en pratique. Sur des exemples d'applications en tomographie X (reconstruction de volume 1024**3 voxels) et en déconvolution de signaux 1D (enregistrement sur plusieurs années de données spectrales de Mars) ou 2D...Aller à la page de la contribution
-
M. Joel Falcou (LRI)30/03/2015 14:25GPGPus and other form of accelerators are becoming a mainstream asset for high performance computing. Raising the programmability of such hardware is paramount to enable the maximum amount of users to discover, master and subsequently use accelerators in there day-to-day research activities. This presentation showcase NT2 - the Numerical Template Toolbox - which is a C++ HPC library...Aller à la page de la contribution
-
M. Postma Bart (INRIA - team AVIZ)30/03/2015 14:35A fundamental way to analyze the brain is by studying brain connectivity, i.e. how brain regions are connected to each other. Two types of brain connectivity exist, anatomical connectivity and functional connectivity, each with their advantages and disadvantages. Being able to study both types of connectivity in concert in real time addresses a need of neuroscientists and neurosurgeons. To...Aller à la page de la contribution
-
M. Florian Beaudette (LLR)30/03/2015 14:45The Matrix Element Method (MEM) is a powerful approach in particle physics to extract maximal information of the events arising from the LHC pp collisions and is currently being deployed in the Higgs->tautau Vector Boson Fusion channel. Compared to other methods requiring trainings, the MEM allows direct comparisons between a theory and the observation. Since this method implies an integration...Aller à la page de la contribution
-
M. Samuel Vaiter (CMAP)30/03/2015 14:55The total variation is used in many applications including imaging, signal processing and machine learning. We developed a distributed algorithm to compute the so-called proximal operator of this regularization. In this talk, I will give some insights on the distribution scheme used to implement the underlying CUDA code. Some applications will be covered, such as 2D/3D denoising and 2D...Aller à la page de la contribution
-
M. Bogdan-Ionut Cirstea (Telecom-ParisTech)30/03/2015 15:35Handwriting recognition is a classical AI problem, which has been studied for around 50 years; in its most recent variant, it deals with the recognition of handwritten lines of text. Beyond its inherent importance, handwriting recognition has also served as a testbed for the introduction of some widely used machine learning algorithms, such as the convolutional neural network (CNN) and the...Aller à la page de la contribution
-
Dr Isabelle Guyon (ChaLearn)30/03/2015 15:45We have been organizing in the recent years a number of machine learning challenges with datasets of increasingly large sizes. Particularly demanding are the computer vision and medical imaging tasks. As challenges in machine learning move into the era of big data, it becomes less and less realistic to move data around to let participants enter challenges. Rather, we promote the use of...Aller à la page de la contribution
-
Prof. Michalis Vazirgiannis (LIX Ecole Polytechnqiue)30/03/2015 15:55Understanding and controlling spreading dynamics in networks assumes identification of the most influential nodes that will trigger efficient information diffusion. It has been shown that the best spreaders are the ones located in the k-core of the network rather than those with the highest degree or centrality [Kitsak et al., Nature Physics 6, 888–893 (2010)]. In this paper, we further...Aller à la page de la contribution
-
30/03/2015 16:05
Choisissez le fuseau horaire
Le fuseau horaire de votre profil: