Obiettivo "Computers are now able to recognize people, to tell a dog from a cat, or to process speech so efficiently that they can answer complicated questions. This was still impossible only a decade ago. This progress is largely due to the development of the artificial “deep-learned neural networks”. Nowadays, “deep learning” is revolutionizing our life, prompting an economic battle between internet giants and the creation of a myriad of start-ups. As attractive and performant as it is, however, many agree that deep learning is largely an empirical field that lacks a theoretical understanding of its capacity and limitations. The algorithms used to ""train"" these networks explore a very complex and non-convex energy landscape that eludes most of the present theoretical methodology in machine learning. The behavior of the dynamics in such complicated ""glassy"" landscape is, however, similar to those that have been studied for decades inthe physics of disordered systems such as molecular and spin glasses.In this project we pursue this analogy and use advanced methods of disordered systems to develop a statistical mechanics approach to deep neural networks. Our first main objective is to create a model for learning features from data via a multi-level neural network. We then regard this model as a kind of a spin glass system amenable to an exact asymptotic analysis via the replica and cavity method. Analyzing its phase diagram and phase transitions shall bring theoretical understanding of the principles behind the empirical success of deep neural networks. This approach will also lead to our second objective: the creation of a new class of fast, efficient, and asymptotically optimal message passing algorithms for deep learning. It is the synergy between the theoretical statistical physics approach and scientific questions from computerscience that makes the project’s objectives feasible and enables a leap forward in our understanding of learning from data." Campo scientifico natural sciencescomputer and information sciencesinternetengineering and technologymaterials engineeringnatural sciencesphysical sciencesclassical mechanicsstatistical mechanicsnatural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learningnatural sciencescomputer and information sciencesartificial intelligencecomputational intelligence Parole chiave spin glasses artificial neural networks Bethe approximation replica method cavity method message passing algorithms belief propagation unsupervised learning graphical models feature learning Programma(i) H2020-EU.1.1. - EXCELLENT SCIENCE - European Research Council (ERC) Main Programme Argomento(i) ERC-2016-STG - ERC Starting Grant Invito a presentare proposte ERC-2016-STG Vedi altri progetti per questo bando Meccanismo di finanziamento ERC-STG - Starting Grant Istituzione ospitante ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE Contribution nette de l'UE € 599 753,36 Indirizzo BATIMENT CE 3316 STATION 1 1015 Lausanne Svizzera Mostra sulla mappa Regione Schweiz/Suisse/Svizzera Région lémanique Vaud Tipo di attività Higher or Secondary Education Establishments Collegamenti Contatta l’organizzazione Opens in new window Sito web Opens in new window Partecipazione a programmi di R&I dell'UE Opens in new window Rete di collaborazione HORIZON Opens in new window Costo totale € 599 753,36 Beneficiari (2) Classifica in ordine alfabetico Classifica per Contributo netto dell'UE Espandi tutto Riduci tutto ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE Svizzera Contribution nette de l'UE € 599 753,36 Indirizzo BATIMENT CE 3316 STATION 1 1015 Lausanne Mostra sulla mappa Regione Schweiz/Suisse/Svizzera Région lémanique Vaud Tipo di attività Higher or Secondary Education Establishments Collegamenti Contatta l’organizzazione Opens in new window Sito web Opens in new window Partecipazione a programmi di R&I dell'UE Opens in new window Rete di collaborazione HORIZON Opens in new window Costo totale € 599 753,36 CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE CNRS Partecipazione conclusa Francia Contribution nette de l'UE € 747 041,64 Indirizzo RUE MICHEL ANGE 3 75794 Paris Mostra sulla mappa Regione Ile-de-France Ile-de-France Paris Tipo di attività Research Organisations Collegamenti Contatta l’organizzazione Opens in new window Sito web Opens in new window Partecipazione a programmi di R&I dell'UE Opens in new window Rete di collaborazione HORIZON Opens in new window Costo totale € 747 041,64