This course develops an up-to-date analysis of machine learning and data analysis from the mathematical and experimental viewpoints. Its topics will be selected in collaboration with the students and will focus on the most relevant and recent mathematical contributions to the understanding of machine learning: mathematical models of optimal networks, performance limits, probabilistic interpretation and Bayesian models. The usage of deep learning will be the main focus. Students will be invited to propose topics and papers of their current research interest, in particular applications to signals, images or other data types, so that participation to the course will foster their own research and in depth understanding of their own research topic. Classic signal and image and data analysis and their interaction with deep learning will be also discussed on demand, for example image denoising and deblurring, motion analysis, image comparison, anomaly detection, object classification, image or signal based medical diagnosis etc.  Since the goal of the course is to help students develop their own research, they will be invited to select one paper or several papers of their particular interest and will commit themselves to an oral exposition and to a written report. If the report reaches sufficient level, publication at IPOL (www.ipol.im) will be considered or submission to a conference or journal will be considered and the instructor will help the student to complete it after the course.  The main evaluation criterion is the production of a good public exposition and a good essay on the topic of interest for the student.

Continuous Assessment: 100%