Machines car learn. Yes, but how? In this talk will be presented a particular kind of neural networks first introduced by G. Hinton in 2006 : the Deep Beliefs Networks. What are the main ideas behind this machine learning algorithm? How can statistics allow to learn abstract representations? Is there any result using this neural network? The aim of this talk is not to study in-depth the mathematical aspect of this algorithm, but rather to focus on the possibilities and the ideas that Deep Beliefs Networks represent.
Denis Maurel is a last year student at Telecom SudParis in France and a current trainee at the University of Vigo. After a preparatory class in Mathematics and Physics and two years in an engineering school, he followed a Specialised Master in Data Analysis that made him get interested in machine learning, especially in neural networks and genetic algorithms.