This presentation covers case-based and model-based reasoning for artificial intelligence. Topics covered are as follows: case-based reasoning, case-based reasoning components; case base, retriever, adapter, refiner, executor, and evalutator; and model-based reasoning.
1. Case-based Reasoning (CBR)
• Collection of a set of cases
– Store a case in the Case Base
• CBR is based on human information processing (HIP)
model in some problem areas
– Thinking about how human processes information
– Try to remember previous case/recall similar cases
& modify to fit a new situation
• Examples: Law, diagnosis, strategic planning
– CEO: How should we modify last year’s plan?
• Human experts depend heavily on past experiences
when solving new problems
2. Case-based Reasoning
• A CBR emulates this HIP model and
maintains a historical case
• It retrieves cases relevant to the present
problem situation from the case base
and decides on the solution to the
current problem on the basis of the
outcomes from previous cases
3. CBR Components
• A case-based ES consists of
– a case base
– a retriever
– an adapter
– a refiner
– an executer
– an evaluator
– (See diagram on next slide)
4. CBR Components
User Interface
Request Relevant
Adapter
Evaluator
Retriever
Case Base
New
Cases
(w/ Results)
Executer Refiner
Prior Cases
Prior
Result
Draft
Solution
Refined
Solution
Solution
Performance
Cases
w/o Results
5. Case-based Parts
• A case base functions as a repository of
prior cases
• The cases are indexed ( a key as with a
database) so that they can be quickly
recalled when necessary
• A case contains the general
descriptions of old problems
• In Knowledge Base: Set of Rules
• In Case Base: Set of Cases
– Creates extra difficulty in retrieving
6. Case-base Parts - The Retriever
• When a new problem is entered into a
case based system, a retriever decides
on the features similar to the stored
cases
• Retrieval is done by using features of
the new cases as indexes into the case
base
7. Case-based Parts - The Adapter
• An adapter examines the differences
between these cases and the current
problem
• It then applies rules to modify the old
solution to fit the new problem
8. Case-based Parts - The Refiner
• A refiner critiques the adapted solution
against prior outcomes
• One way to do this is to compare it to
similar solutions of prior cases
• If a known failure exists for a derived
solution, the system then decides
whether the similarities is sufficient to
suspect that the new solution will fail
9. Case-based Parts - The Executor
• Once a solution is critiqued, an executer
applies the refined solution to the
current problem
10. Case-based Parts - The Evaluator
• If the results are as expected, no further
analysis is made, and the cases and its
solution is stored for use in future
problem solving
• If not, the solution is repaired
11. Model-based Reasoning
• A model-based system is based on a model of the
structure and behavior of the device that the
system is designed to simulate
• Used for well structured problems
– Not for stock pricing/modeling, not well
structured
– Engineering Problems
• Ex: Diagnosing hardware or a machine
• Ex: Automobile diagnostics
• Based on written documentation
• The problem is extracting knowledge
12. Model-based Reasoning
• Observed behavior (what the device is
actually doing) is compared with
predicted behavior (what the device
should do)
• The difference between them is called a
discrepancy, indicating a defect
• Then a process is initiated to diagnose
the nature and location of the defect
• Could be a mathematical equation
14. Model-based Reasoning
• Correct Operation:
• Assume we have a model-based system built to diagnose the
following simple device with 3 multipliers and 2 adders
• Once the logic is developed, executes quickly
A = 3
B = 3
C = 2
D = 2
A = 3
Mult - 1
Mult - 2
Mult -3
Add - 3
Add - 3
F = 12
G = 12
3
2
3
2
2
3
6
6
6
6
6
6
6
12
12
15. Model-based Reasoning
• Incorrect Operation:
– Diagnostics
A = 3
B = 3
C = 2
D = 2
A = 3
Mult - 1
Mult - 2
Mult -3
Add - 3
Add - 3
F = 10
G = 12
3
2
3
2
2
3
6
6
6
6
6
6
6
10
12
6
6
6
Culprit of problem?
No - 6 comes out
Culprit of problem?
No - 6 comes out
Culprit of problem?
No - 6 comes out
Culprit of problem?
Yes - 10 comes out,
problem with Carry Bit
Predicate Logic
can be used here