Case Based Reasoning

1
Case Based Reasoning

• Lecture 5 Reuse, Adaptation and Retention

2
Outline

• Re-use
• How to re-use retrieved solutions
• Adaptation
• Why might we want to revise the solution?
• Types of adaptation
• Retention
• Why might we wish to retain cases?

3
Re-Using Retrieved Solutions

• Single retrieved solution
• Re-use this solution
• Multiple retrieved solutions
• Vote/average of retrieved solutions
• Weighted according to
• Ranking
• Similarity
• Iterative retrieval
• Solve components of the solution one at a time

4
Multiple Retrievals

• Whole solution generated in single retrieval
• Single components generated in each retrieval
• Parallel
• Incremental

5
When is Adaptation Needed?

• Classification
• All solutions likely to be represented in
  case-base
• Adaptation corrects for lack of cases
• Constructive problem solving
• All designs unlikely to be represented in
  case-base
• Retrieved cases suggest initial design
• Adaptation alters the design to reflect novel
  feature values

6
Assumptions for Adaptation

• Similar problems have similar solutions
• The effort required to adapt a retrieved solution
  will be less the more similar it is to the
  required solution

7
How to Adapt the Solution

• Adaptation alters proposed solution
• takes account of differences between new and
  retrieved problems
• Null adaptation - copy retrieved solution
• Used by CBR-Lite systems
• Manual or interactive adaptation
• User adapts the retrieved solution (Adapting is
  easier than solving?)
• Automated adaptation
• CBR system is able to adapt the retrieved
  solution
• Adaptation knowledge required

8
Automated Adaptation Methods

• Substitution
• change some part(s) of the retrieved solution
• simplest and most common form of adaptation
• Transformation
• alters the structure of the solution
• Generative
• replays the method of deriving the retrieved
  solution on the new problem
• method of solution is part of retrieved case
• most complex form of adaptation

9
Examples of Adaptation

• CHEF
• CBR system to plan Szechuan recipes
• Hammond (1990)
• Substitution adaptation
• substitute ingredients in the retrieved recipe to
  match the menu
• Retrieved recipe contains beef and broccoli
• New menu requires chicken and snowpeas
• Replace chicken for beef, snowpeas for broccoli
• Transformation adaptation
• Add, change or remove steps in the recipe
• Skinning step added for chicken, not done for
  beef

10
Examples of Adaptation

• Car diagnosis example
• Symptoms, faults and repairs for brake lights are
  analogous to those for headlight
• Substitution brake light fuse
• Planning example
• Train journeys and flights are analogous
• Transformation flights need check-in step added

11
Adaptation in CBR-Works

• Provides adaptation rules
• IF a THEN b
• classic production rules
• Example
• Add 1000 to the price of a new car for a
  different colour

12
Recalculate price for new colour

• ? QueryColour isRegular
• ? RetrievedColour isRegular
• ? QueryColourltgtRetrievedColour
• ? ?OldPrice RetrievedPrice
• ? ?OldPrice be_of_type Integer
• ? ?NewPrice ?OldPrice 1000
• ? ?NewPrice be_of_type Integer
• ! ResultPrice ?NewPrice
• ! ResultColour QueryColour

Conditions
Actions
13
Adaptation in CBR-Works Example

• Retrieval without adaptation

14
Adaptation in CBR-Works Example

• Retrieval adaptation
• Predicting value of the price attribute

15
Adaptation in CBR-Works Example

• Adaptation rule to predict the value of Price

16
Other Rules in CBR-Works

• CBR-Works also uses completion rules to
• calculate a dependent attribute value
• set default value
• alter the feature weights in certain
  circumstances
• Used to
• complete a query
• fill-in missing data during case creation
• alter similarity calculations for retrieval

17
Adaptation in ReCall

• By default, ReCall uses the vote mechanism of
  k-NN to predict a value for the target attribute.
  
• E.g., the predicted value of the query (shown
  here in black) will be grey according to a 3-NN
  algorithm which retrieves 3 similar cases (2 in
  grey and 1 in beige)

18
Adaptation in ReCall

• Alternatively, ReCall allows you to write
  adaptation rules to predict a value for your
  query based on a single (most similar) case.
• You can use ReCalls own language, or use the
  more powerful and widely used language Tcl.
• To find out more refer to ReCalls Lab notes.

19
Two Schools of Thought in CBR

• Adaptation is the most contentious issue in CBR
• One group believes adaptation is not important
• The problem cannot be solved using CBR
• A CBR system without adaptation capabilities is
  called CBR Retrieval System
• Others believe it is vital
• Without adaptation and generation of new
  solutions there is no reasoning in CBR
• A CBR system with adaptation capabilities is
  called fully-fledged CBR system

20
Retention

• What can be learned
• New experience to be retained as new case
• Representing the new case
• Contents of new case
• Indexing of new case
• Forgetting cases
• For efficiency or because out of date
• Deleting an old case
• Old is not necessarily bad
• Does it leave a gap?

21
Example

• OutlookCloudy Temp.Cool HumidityHigh WindyFals
  e Play Yes
• OutlookCloudy Temp.Mild HumidityHigh WindyFals
  e Play No
• Do we need to retain the new case?
• Do we need to rebuild the decision tree index?

22
Summary

• Reuse
• Initial solution from retrieved cases
• Revise
• Adapt initial solution to reflect differences
  between new and retrieved problems
• CBR-Works adaptation rules
• Retain
• When to retain and whether to replace
• Representation and indexing

23
Reading

• Research Papers
• S. Craw, J. Jarmulak R. Rowe. Learning and
  Applying Case-Based Adaptation Knowledge.
  Proceedings 4th ICCBR Conference, p131-145, 2001.
  www.comp.rgu.ac.uk/staff/smc/papers/iccbr01smc.pdf
  
• B. Smyth M. T. Keane. Adaptation-Guided
  Retrieval Questioning the similarity assumption.
  Artificial Intelligence 102249-293, 1998.
  www.cs.ucd.ie/staff/mkeane/SmythKeane98.pdf