arrow drop search cross

Deep Neural Networks and Gaussian Processes for temporal analysis of clinical data

Deep Neural Networks (DNN) and Gaussian Processes (GP) are modern machine learning methods that have been shown to outperform standard analysis approaches in many applications


18/04/2019   :   10h00
Salle des conferences MSI Sophia Antipolis
 Marco Lorenzi (Inria Sophia Antipolis Méditerranée)
Publication : 18/04/2019
Partager cet article :

Abstract: Deep Neural Networks (DNN) and  Gaussian Processes (GP) are modern machine learning methods that have been shown to outperform standard analysis approaches in many applications. These approaches come with theoretical guarantees of being able to approximate any function, thus providing high flexibility and expressivity in inferring complex data relationships. Furthermore, recent methodological developments have allowed to reformulate these learning frameworks to efficiently account for uncertainty quantification in complex and noisy data. This talk will focus on the recent advancements in GP and probabilistic DNN modeling, and will illustrate novel applications of these methodologies to the analysis of biomedical data. In particular, we will provide examples of spatio-temporal modeling and source separation of high-dimensional brain imaging data, disease progression modeling through time warping of time-series, and bio-mechanical modeling via derivative-constrained GP.

Please consult the MSI's agenda to find out about forthcoming seminars and other scientific events