|
|
|
Finite State Automata for Probabilistic Artificial Intelligence |
|
|
12h L / 3h R / 1 WE + 1 OE / 4 ECTS credits in common with IIC_OAD1 and IIC_OAD3 / IIC_OAD2 |
|
|
Olivier PIETQUIN |
|
|
This course will take up the main graph methods used in probabilistic AI. Specifically, we will talk about standard and dynamic bayesian networks,about hidden Markov models and about standard and partially observable Markov decision process. These techniques have applications in the fields of decision making, data fusion, pattern recognition and many others too. The course will also try to give a unified view of those techniques.
|
|
|
Decision Graphs and Bayesian Networks |
|
|
In this part, we will tackle the standard and dynamic bayesian networks, and also decision graphs. Statistical inference methods will be described together with network learning techniques, from the parametrical and topological points of view. Applications to decision making, data fusion, optimal control and behavioural modeling will be presented.
|
|
|
Hidden Markov Models |
|
|
In this part, we will present learning and resolution algorithms (Viterbi, Baum-Welch algorithms) for Hidden Markov Models (HMM). These models are very useful for pattern recognition with temporal or spatial distortion. They have been used for vocal recognition since the end of the 70’s and are still today at the heart of all voice recognition commercial systems. Gesture or speaker recognition will also be considered.
|
|
|
Markov Decision Process |
|
|
Observable Markov Decision Process (MDP) and Partially Observable Markov Decision Process (POMDP) will be described in this part of the course. Thes methods are used for optimal control when it is impossible to supervise the system but reinforcement learning is required at the same time. Main reinforcement learning methods for MDP and POMDP will be presented. Applications to robotics and to human-computer interaction optimization will be considered.
|
|
|
|
|
References
J. Pearl. Probabilistic Reasoning in Intelligent Systems. Morgan Kaufman, San Mateo, California, 1988.
F.V. Jensen. Bayesian Networks and Decision Graphs. Springer-Verlag, 2000
K.B. Korb, A.E. Nicholson. Bayesian Artificial Intelligence. Chapman & Hall/CRC, 2004.
L. Rabiner and B.-H. Juang. Fundamentals of Speech Recognition. Prentice Hall Signal Processing Series, 1993.
R. J. Elliott, L. Aggoun, J.B. Moore. Hidden Markov Models : Estimation and Control. Springer, 1997.
R.S. Sutton and A.G. Barto. Reinforcement Learning. An Introduction. Cambridge, MA: MIT Press, 1998.
E.J. Sondik. The optimal control of partially observable Markov processes over the infinite horizon. Operations Research 26(2), 283--304 (March-April 1973). |
|
|
Last update 06/07/2007 by Cl.M. |
|