Course description: Our information society produces an ever-increasing flow of unstructured data of various types (texts, audio, image, video, etc) that needs to be dealt with quickly and effectively. In the face of such polymorphic data, probabilistic models have emerged through their ability to digest the variability of information into effective and sound statistical models. In the last decades, these models have become indispensable tools for information management and decision-making. The course is divided into two main parts. The first part deals with the basic concepts and their computational manipulation: directed and undirected graphical models, and the associated algorithms. In the second part, we focus more specifically on (a) the estimation of latent variable models: (b) approximate inference. We will illustrate these methods with applications from the text mining literature (text classification and clustering, question answering, sentiment analysis, etc).
Main themes :
- Directed graphical model and probabilistic reasoning
- Undirected graphical model
- Exact inference in graphical models
- EM and latent variable models
- Approximate inference: variational techniques
- Approximate inference: sampling techniques
Language: English
Numerus Clausus: 24
Recommended readings:
- Probabilistic Graphical Models: Principles and Techniques by Daphne Koller and Nir Friedman. MIT Press.
- Pattern Recognition and Machine Learning by Chris Bishop.
- Machine Learning: a Probabilistic Perspective by Kevin P. Murphy. MIT Press
- Modeling and Reasoning with Bayesian networks by Adnan Darwiche. Information Theory, Inference, and Learning Algorithms by David J. C. Mackay. [Available online.]
- Graphical models, exponential families, and variational inference by Martin J. Wainwright and Michael I. Jordan. [Available online]
Prerequisites: Basic statistics and optimization
Grading: Final exam