Intelligent Systems CSCI 6501 Dr. D. Riordan

1
Intelligent SystemsCSCI 6501Dr. D. Riordan

• Pradeep Monga (B00342080)
• Satwant Sandhu (B00201045)

2
Project Proposal

Structure for Credit-Apportionment Process in
Rule-Based Systems
3
Agenda

• Problem Definition
• More Background
• Algorithms
• Framework
• Application
• Task Allocation
• References

4
Problem Definition

• Usually, rule based systems gain experience
  through feedback from environment. In many cases
  these feedbacks are in the form of numerical
  values indicating success or failure of
  problem-solving process. Such numerical valued
  feedback is called payoff. The problem of
  apportioning these payoffs (representing credit
  or blame) to the rules in a rule based system is
  called the credit-apportionment problem.

5
More Background

• Introduction about rule based systems
• Three essential Parts-
• Production memory
• Working Memory
• An Inference Engine

6
Contd...

• Inference engine performs Recognition act cycle-
• Match process
• Conflict Resolution Process
• Act process

7
Contd

• In rule based systems, rules are assigned scalar
  valued strengths representing their past
  usefulness.
• These strengths provide useful basis for learning
  activities
• Refine rules by adjusting their strengths
• Modify old rules using strengths as biases

8
Algorithms

• The strengths are adjusted by various types of
  Credit-Apportionment algorithms.
• The Bucket Brigade algorithm designed for
  conducting Credit-Apportionment process in the
  classifier system is one of them.

9
Framework

• The framework proposed here, is important in
  studying the Credit-Apportionment process because
  it provides a formal basis for the problem
  analysis and algorithm design. The main
  components of the framework include-
• System Environment Model
• Principles of Usefulness

10
Contd

• A system environment model integrates the effects
  of payoffs with other relevant parts of a rule
  based system to model all the different internal
  and external activities of the rule based system.
• Principles of usefulness define the inherent
  usefulness of rule actions and provide the
  semantic aspect of the credit - apportionment
  process.

11
Scope

• Here we are concentrating on finding a good
  solution and not an optimum one. By a good
  solution we mean to drive the environment to a
  terminal state with high payoff. The optimal way
  is to drive the environment to a terminal state
  with a minimum number of rules firing or with
  minimum number of environment states traversed.

12
Application

• This credit apportionment problem has been made
  use of in designing an expert system called -
• GAMBLE
• (Genetic Algorithm Based Machine
• Learning Expert)

13
Abstract detail

• The present day society provides a working
  environment that is insensitive to human
  behavior. From a macroscopic viewpoint,
  educational programs are run in largely
  open-loop fashion. In this project, we examine
  the entrance procedure for a student seeking
  admission to an engineering institute. GAMBLE is
  a simple classifier system, which learns from the
  performance of the previous batches. This
  experience coupled with information about
  students aptitude enables the expert to guide
  the student towards the branch best suited for
  him.

14
Task Allocation

• Research on Rule based systems Satwant
• Research on applications of
• Credit-apportionment process Pradeep
• Implementation of the project
• and documentation Jointly

15
REFERENCES

• http//www.icce2001.org/cd/pdf/p14/IN002.pdf
• GAMBLE expert system developed using Credit
  apportionment process and Bucket Brigade
  Algorithm at Indian Institute of Technology,
  Roorkee India 2001.
• Framework for the Credit-Apportionment Process in
  Rule-Based Systems,Huang, D. Systems, Man and
  Cybernetics, IEEE Transactions onOn page(s)
  489-498 Volume 19,   Issue 3,   May/Jun 1989
• Bucket Brigade Performance 1 long sequences of
  classifiers in Genetic algorithms and their
  application proc 2nd int. conf on GA.
  J.Grefenstette, Ed. July 1987
• A study on apportionment of credits of fuzzy
  classifier system for knowledge acquisition of
  large scale systems Nakaoka, K. Furuhashi, T.
  Uchikawa, Y. Fuzzy Systems, 1994. IEEE World
  Congress on Computational Intelligence.,
  Proceedings of the Third IEEE Conference on ,
  26-29 June 1994
• Fast and feasible reinforcement learning
  algorithm, Ono, S.   Inagaki, Y.   Aisu, H.  
  Sugie, H.   Unemi, T. Fuzzy Systems, 1995.
  International Joint Conference of the Fourth IEEE
  International Conference on Fuzzy Systems and The
  Second International Fuzzy Engineering
  Symposium., Proceedings of 1995 IEEE
  International Conference on03/20/1995
  -03/24/1995,  20-24 Mar 1995  

16
ANY COMMENTS !!