Course : Artificial Intelligence (COMP6065)

Non-official Slides

Knowledge in Learning

Session 20

Revised by Williem, S. Kom., Ph.D.

1
 Learning Outcomes
At the end of this session, students will be able to:

• LO 5 : Apply various techniques to an agent when acting under

certainty

• LO 6 : Apply how to process natural language and other

perceptual signs in order that an agent can interact
intelligently with the world

2
 Outline
1. A Logical Formulation of Learning

2. Knowledge in Learning

3. Explanation Based Learning

4. Learning Using Relevance Information

5. Inductive Logic Programming

6. Summary

3
 A Logical Formulation of Learning
• In this session, the hypotheses are represented by a set of
logical sentences

– Allows for incremental construction of hypotheses

– Allows for prior knowledge

• Example

– The restaurant learning problem can be described as

4
 A Logical Formulation of Learning
• Study case

– Change the example in logical representation

5
 A Logical Formulation of Learning
• Thus, a decision tree can be
interpreted in logical
representation

6
 A Logical Formulation of Learning
• As the example arrive, hypotheses that are not consistent with
the examples can be ruled out

• What is the meaning when an example is consistent?

– False negative: When the hypothesis says that it is

negative, but in fact it is positive

– False positive: When the hypothesis says that it is positive,

but in fact it is negative

7
A Logical Formulation of Learning

Hypothesis False negative False positive

8
 A Logical Formulation of Learning
• How to find the logically consistent hypothesis?

– Current-best-hypothesis search

• To maintain a single hypothesis, and to adjust it as new

examples arrive

– Least-commitment search

• Candidate elimination algorithm has an incremental

property, where one never has to god back and
reexamine the old examples

9
 Current-best-hypothesis Search
• Suppose we have some hypothesis hr

– As long as each new example is consistent, we do nothing

– If there is a false negative example?

• Generalization: We extend the hypothesis region

– If there is a false positive example?

• Specialization: We reduce the hypothesis region

10
 Current-best-hypothesis Search
(extensions of predictor Hr)

Initial False False

hypothesis negative positive
a generalization a specialization
Generalization e.g. via dropping conditions
Alternate(x)Patrons(x, Some)  Patrons(x, Some)

Specialization e.g. via adding conditions or via removing disjuncts

Alternate(x)  Alternate(x)Patrons(x, Some)
11
 Least-commitment search
• Assuming the original hypothesis space is the right answer, the
reduced disjunction must still contain the right answer because
only incorrect hypotheses have been removed

• The set of hypotheses remaining are called version space

• The learning algorithm is called version space learning or

candidate elimination algorithm

• The problem is how to deal with all remaining hypotheses?

– HUGE!

12
 Least-commitment search
• The problem is similar to how we represent all numbers between
1 and 2

– Using boundary!!

• The boundary in the version space is generalization/specialization

– A most general boundary (G-set)

– A most specific boundary (S-set)

• Everything in between is guaranteed to be consistent with the

examples

13
Version Space

14
 Version Space Learning
• Algorithm:

– False positive for Si: Si is too general, and there are no

consistent specializations for Si, so throw Si out of S-Set

– False negative for Si: Si is too specific, so replace it with

all its immediate generalizations.

– False positive for Gi: Gi is too general, so replace it with

all its immediate specializations.

– False negative for Gi: Gi is too specific, but there are no

consistent generalizations of Gi, so throw Gi out of G-Set
15
 Version Space Learning
• We continue for each example until:

– We have exactly one hypothesis left

– The version space collapses – either S-set or G-set empty

– We run out of examples and have several hypotheses

remaining

16
 Candidate Elimination Algorithm
• Given : A representation language and a set of positive
and negative examples expressed in that language.

• Compute : A concept description that is consistent with all

the positive examples and none of the negative examples.

1. Initialize G to contain one element : the null description (all

features are variables)

2. Initialize S to contain one element : the first positive

example

17
 Candidate Elimination Algorithm
3. Accept a new training example.

• If it is a positive example

– Remove from G any descriptions that do not cover the

example

– Update the S set to contain the most specific set of

descriptions in the version space that cover the example
and the current element of the S set

18
 Candidate Elimination Algorithm
• If it is a negative example
– Remove from S any descriptions that cover the example
– Update the G set to contain the most general set of
descriptions in the version space that do not cover the
example
• If S and G are both singleton sets, then if they are identical,
output their value and halt.
• If they are both singleton sets but they are different, then the
training cases were inconsistent. Output this result and halt.
• Otherwise, go to step 3.
19
 Candidate Elimination
Algorithm
G : { obj(X,Y,Z) }, S: { } Pos : obj(small,red,ball)

Initialize G to be the most general concept in the space, and

Initialize S to the first pos. training instance

G : { obj(X,Y,Z) }, S :{ obj(small, red, ball) }

20
 Candidate Elimination
Algorithm
G : { obj(X,Y,Z) }, S: { } Pos : obj(small,red,ball)
Initialize G to be the most general concept in the space, and
Initialize S to the first pos. training instance
G : { obj(X,Y,Z) }, S :{ obj(small, red, ball) } Neg : obj(small,blue,ball)

Delete all members of S that match n

For each g that match n, replace g with its most general specialization
that do not match n

G : { obj(X,red,Z) }, S :{ obj(small, red, ball) }

21
 Candidate Elimination
Algorithm
G : { obj(X,Y,Z) }, S: { } Pos : obj(small,red,ball)
Initialize G to be the most general concept in the space, and
Initialize S to the first pos. training instance
G : { obj(X,Y,Z) }, S :{ obj(small, red, ball) } Neg : obj(small,blue,ball)

Delete all members of S that match n

For each g that match n, replace g with its most general specialization
that do not match n

G : { obj(X,red,Z) }, S :{ obj(small, red, ball) } Pos : obj(large,red,ball)

Delete all members of G that fail to match p

For every s, if s does not match p, replace s with its most specific
generalization that match p

G : { obj(X,red,Z) }, S :{ obj(X, red, ball) }

22
 Candidate Elimination
Algorithm
G : { obj(X,Y,Z) }, S: { } Pos : obj(small,red,ball)
Initialize G to be the most general concept in the space, and
Initialize S to the first pos. training instance
G : { obj(X,Y,Z) }, S :{ obj(small, red, ball) } Neg : obj(small,blue,ball)

Delete all members of S that match n

For each g that match n, replace g with its most general specialization
that do not match n

G : { obj(X,red,Z) }, S :{ obj(small, red, ball) } Pos : obj(large,red,ball)

Delete all members of G that fail to match p

For every s, if s does not match p, replace s with its most specific
generalization that match p

G : { obj(X,red,Z) }, S :{ obj(X, red, ball) } Neg: obj(large,red,cube)

Delete all members of S that match n

For each g that match n, replace g with its most general specialization
that do not match n
23
G : { obj(X,red,ball) }, S :{ obj(X, red, ball) }
 Candidate Elimination
Algorithm
Positive and Negative Examples of the Concept “Japanese economy car”

origin : Japan origin : Japan origin : Japan

mfr : Honda mfr : Toyota mfr : Toyota
color : Blue color : Green color : Blue
decade : 1980 decade : 1970 decade : 1990
type : Economy type : Sports type : Economy
(+) (-) (+)
origin : USA origin : Japan
mfr : Chrysler mfr : Honda
color : Red color : White
decade : 1980 decade : 1980
type : Economy type : Economy
(-) (+)

24
 Candidate Elimination
Algorithm
A Version Space Example
• G = {(x1,x2,x3,x4,x5)}
S = {(Japan, Honda, Blue, 1980, Economy)}

• G = {(x1, Honda,x3,x4,x5) ,(x1,x2,Blue,x4,x5),

(x1,x2,x3,1980,x5),(x1,x2,x3,x4,Economy)}
• S = {(Japan, Honda, Blue, 1980, Economy)}

• G = {(x1,x2, Blue, x4,x5), (x1,x2,x3,x4,Economy)}

• S = {(Japan, x2, Blue, x4, Economy)}

• G = {(x1,x2, Blue, x4,x5), (x1,x2, Blue,x4,Economy),

(Japan,x2,x3,x4, Economy)}
• S = {(Japan,x2, Blue,x4, Economy)}

• G = {(Japan,x2,x3,x4,Economy)}
• S = {(Japan,x2,x3,x4,Economy)} 25
 Knowledge in Learning
• Let Descriptions denote the conjunction of all the example
descriptions in the training set, and let Classifications denote
the conjunction of all the example classification.

• Then a Hypothesis that "explains the observations" must

satisfy the following property (recall that |= means "logically
entails"):

Hypothesis ۸ Descriptions |= Classifications

26
Knowledge in Learning

27
 Knowledge in Learning
• There are three kinds of learning using prior knowledge:

– Explanation-based learning (EBL)

– Relevance-based learning (RBL)

– Knowledge-based inductive learning (KBIL)

• Inductive logic programming (ILP)

28
 Explanation Based Learning
• A method for extracting general rules from individual
observations

• As example, consider the problem of differentiation!

– X2 = 2X

• Compare the student with and without calculus

knowledge

• Which one is faster?

29
 Explanation Based Learning
• The technique of memoization has long been used in
computer science to speed up programs by saving the results
of computation.

• The basic idea of memo functions is to accumulate a database

of input—output pairs; when the function is called, it first
checks the database to see whether it can avoid solving the
problem from scratch.

• Explanation –based learning takes this a good deal further, by

creating general rules that cover an entire class of cases.
30
 Learning Using Relevance Information
• A method for extracting general rules from the relevance
between prior knowledge and observations to explain the
observations

• RBL does not produce hypotheses that go beyond the logical

content of the background knowledge

– Cannot create a new knowledge from scratch

31
 Learning Using Relevance Information
• Functional dependencies or determinations occur so
commonly in certain kinds of applications

• Determinations specify a sufficient basis vocabulary from which

to construct hypotheses concerning the target predicate

32
 Inductive Logic Programming
• A method for extracting general rules from the background
knowledge and new hypothesis to explain the examples

• Inductive logic programming (ILP) combines inductive

methods with the power of first-order representations,
concentrating in particular on the representation of
hypotheses as logic programs.

33
 Inductive Logic Programming
• It has gained popularity for three reasons :
1. ILP offers a rigorous approach to the general knowledge-
based inductive learning problem.

2. ILP offers complete algorithms for inducing general, first-

order theories from examples, which can therefore learn
successfully in domains where attribute-based algorithms
are hard to apply.

3. Inductive logic programming produces hypotheses that

are (relatively) easy for humans to read.
34
 Inductive Logic Programming
• Example

– parent_of(charles,george)

– parent_of(george,diana)

– parent_of(bob,harry)

– parent_of(harry,elizabeth)

– grandparent_of(X,Y) : parent_of(X,Z), parent_of(Z,Y)

• Question query Answer

• grandparent_of(X,Y)? • grandparent_of(charles,diana)
• grandparent_of(bob,elizabeth)
35
 Summary
• The use of prior knowledge in learning leads to a picture of
cumulative learning

• Prior knowledge helps learning by

– Eliminating otherwise consistent hypotheses

– Filling in the explanation examples, allowing shorter

hypotheses

36
 References
• Stuart Russell, Peter Norvig. 2010. Artificial Intelligence : A
Modern Approach. Pearson Education. New Jersey.
ISBN:9780132071482

• http://aima.cs.berkeley.edu

37
Version Space

Case Study

38
 Candidate Elimination
Algorithm
Exercises : How About The Concept of Elephant ?

39