This course goes beyond the phase of collecting large volumes of data by focusing on how machine learning algorithms can be rewritten and extended to scale for petabytes of structured and unstructured data. Also sophisticated models for predictions are included. The course is divided into three main parts. The first part deals with the design and development of algorithms allowing the behavior of computers to evolve based on empirical data, such as databases or sensory data. We also define supervised, unsupervised and reinforcement learning. The second part introduces deep learning as well as key network architectures including: convolutional neural networks, autoencoders, recurrent neural networks, long-term short-term networks “LSTM ". This part also covers deep reinforcement learning. The third part deals with the processing of natural languages: Indeed, research in automatic processing of natural languages is a field of artificial intelligence aiming at the development of automated techniques for the manipulation of language data, in textual or sound forms. The immediate applications of these techniques include the development of more natural textual interfaces, the automatic translation of documents, the detection of spam, the search for information in a collection of documents, the systems of questions / answers, and several others. This part introduces the student to the following subjects: Introduction to the problem of automatic processing of natural language and its applications.

Temps présentiel : 35 heures

Charge de travail étudiant : 70 heures

Méthode(s) d'évaluation : Examen final