ME 578 Probabilistic Reasoning                    

(Spring Semester)
Credit Structure:  (3+0) 3

Description :

Introduction to probabilistic reasoning tools that are prevalent in conventional AI topics. Decision making by often incomplete prior probabilities. Methods to handle inconsistent and ambiguous data. Modeling of sensors and noise. Optimizing sensory data by filters in noisy environments. Entropy as a possible decision making tool in connection with information theory. Probabilistic reasoning  is a mathematical way of inference that is already in use as an agent (robot) reasoning, which is the best alternative to rock-rigid deterministic paradigms in a noisy, stochastic agent milieu. 


Course Objectives :

This course is meant to draw attention on the probabilistic side of  decision making in a dynamic world. This course is complimentary in nature to conventional AI topics. Graduate students will be exposed to less known techniques such as rough sets and dempster shafer theory in connection with Bayesians in parameter estimation. The students are expected to grasp the basics in artificial reasoning which could be implemented in his or her research field, irrespective of the specialty.


Prerequisites :

A working knowledge of Matlab,  a working knowledge in at least one another programming language, and a positive attitude in the face of hardships.

Textbooks :

None.

References :

  1. "Probability,Random Variables,and Random Signal Principles", Peyton Z. Peebles, 3rd Ed., 1993 McGraw-Hill.
  2. "Mathematical Approaches to Neural Networks", J.G.Taylor, 1993, North Holland.
  3. "An Introduction to Stochastic Modeling", H.M.Taylor, S. Karlin, 3rd Ed., 1998, Academic Press.
  4. "Introduction to Random Signals and Applied Kalman Filtering", R.G.Brown, P.Y.C. Hwang, 2nd Ed., 1992,  John Wiley&Sons.
  5. "Fundamentals of the New Artificial Intelligence: Beyond Traditional Paradigms", T. Munakata, 1998, Springer-Verlag,.

Syllabus :

1) Introduction (1 wk)
2) Basic Probability (1 wk)
3) Stochastic Calculus (1.5 wks)
4) Parameter Estimation (2 wks)
5) Handling Uncertainty and Inconsistency  (2 wks)
6) Sensor&Noise Modeling (3 wks)
7) Entropy and Information Theory As a Decision Tool (2 wks)

Computer Use :

Homeworks and term papers will require basic coding skills.

List of Term Papers and Guidelines

The List of Selected Paper Topics

1.    Fractal geometry in fault diagnosis, by Aysun Kaya.

2.    Langevin equations for stochastic data analysis in noisy time series, by      Gökhan Bozokalfa.

3.    Study of a population of heterogeneous individuals in GAs, by Selda Alpay, Levent Bilir.

4.    Implementation of on-line learning paradigms based on Statistical Learning Theory, by Murat Karakurt.

5.    Dempster-Shafer Theory in connection with sensor fusion for landmine detection, by Çağdaş Bayram.

 

Conferences, Communities, Networks :

  • EURON, European Robotics Research Network.

 

Lecture Notes

Lecture 1,  Lecture 3,