Machine Learning

CS 165B
Spring 2012

1
 Course outline

• Introduction (Ch. 1)
• Concept learning (Ch. 2)
• Decision trees (Ch. 3)
• Ensemble learning
• Neural Networks (Ch. 4)
• …

2
 Schedule

• Homework 1 due today

• Homework 2 on decision trees will be handed out Thursday
4/19; due Wednesday 5/2
• Project choices by Friday 4/20
– Topic of discussion section

3
Projects
• Projects proposals are due by Friday 4/20.
• 2-person teams
• If you want to define your own project:
– Submit a 1-page proposal with references and ideas
– Needs to have a significant Machine Learning
component
– You may do experimental work, theoretical work, a
combination of both or a critical survey of results in
some specialized topic.
• Originality is not mandatory but is encouraged.
• Try to make it interesting!

4
 Decision tree learning
 Decision tree representation
– Most popular method for representing discrete TE’s
– Decision tree represents disjunction of conjunctions
of attribute values
 More general H-representation than in concept learning

 ID3 learning procedure based on

– Entropy of set of +/- TEs
– Information gain from splitting set with use of attribute
– Greedy, hill-climbing algorithm
 Characterization of ID3 algorithm and its search space
 Overfitting issue
 Generalizations of basic procedure

5
Training Examples

Day Outlook Temp Humidity Wind Tennis?

D1 Sunny Hot High Weak No
D2 Sunny Hot High Strong No
D3 Overcast Hot High Weak Yes
D4 Rain Mild High Weak Yes
D5 Rain Cool Normal Weak Yes
D6 Rain Cool Normal Strong No
D7 Overcast Cool Normal Strong Yes
D8 Sunny Mild High Weak No
D9 Sunny Cool Normal Weak Yes
D10 Rain Mild Normal Weak Yes
D11 Sunny Mild Normal Strong Yes
D12 Overcast Mild High Strong Yes
D13 Overcast Hot Normal Weak Yes
D14 Rain Mild High Strong No
 Decision Trees
• Decision tree to represent learned target functions
– Each internal node tests an attribute
– Each branch corresponds to attribute value
– Each leaf node assigns a classification
Outlook

• Can be represented
sunny overcast rain
by logical formulas
Humidity Yes Wind

high normal strong weak

No Yes No Yes

7
 Representation in decision trees

 Example of representing rule in DT’s:

if outlook = sunny AND humidity = normal
OR
if outlook = overcast
OR
if outlook = rain AND wind = weak
then playtennis

8
 Applications of Decision Trees

 Instances describable by a fixed set of attributes and their values

 Target function is discrete valued
– 2-valued
– N-valued
– But can approximate continuous functions
 Disjunctive hypothesis space
 Possibly noisy training data
– Errors, missing values, …
 Examples:
– Equipment or medical diagnosis
– Credit risk analysis
– Calendar scheduling preferences

9
 Top-Down Construction

 Main loop:
1. Choose the “best” decision attribute (A) for next node
2. Assign A as decision attribute for node
3. For each value of A, create new descendant of node
4. Sort training examples to leaf nodes
5. If training examples perfectly classified, STOP,
Else iterate over new leaf nodes
 Grow tree just deep enough for perfect classification
– If possible (or can approximate at chosen depth)
 Which attribute is best?

10
 Choosing Best Attribute?
• Consider 64 examples, 29+ and 35-
• Which one is better?
29+, 35- A1 29+, 35- A2
t f t f

25+, 5- 4+, 30- 15+, 19- 14+, 16-

• Which is better?
29+, 35- A1 29+, 35- A2
t f t f

21+, 5- 8+, 30- 18+, 33- 11+, 2-

11
Entropy

• A measure for
– uncertainty
– purity
– information content
• Information theory: optimal length code assigns (- log2p) bits to
message having probability p
• S is a sample of training examples
– p+ is the proportion of positive examples in S
– p- is the proportion of negative examples in S
• Entropy of S: average optimal number of bits to encode information
about certainty/uncertainty about S
Entropy(S) = p+(-log2p+) + p-(-log2p-) = -p+log2p+- p-log2p-
• Can be generalized to more than two values
12
 Entropy

 Entropy can also be viewed as measuring

– purity of S,
– uncertainty in S,
– information in S, …
 E.g.: values of entropy for p+=1, p+=0, p+=.5
 Easy generalization to more than binary values
– Sum over pi *(-log2 pi) , i=1,n
 i is + or – for binary
 i varies from 1 to n in the general case

13
 Choosing Best Attribute?
• Consider 64 examples (29+,35-) and compute entropies:
• Which one is better?
E(S)=0.993
29+, 35- A1 E(S)=0.993 29+, 35- A2
t f t f
0.650 0.522 0.989 0.997
25+, 5- 4+, 30- 15+, 19- 14+, 16-

• Which is better?
E(S)=0.993 E(S)=0.993
29+, 35- A1 29+, 35- A2
t f t f
0.708 0.742 0.937 0.619
21+, 5- 8+, 30- 18+, 33- 11+, 2-

14
 Information Gain
• Gain(S,A): reduction in entropy after choosing attr. A
Sv
Gain( S , A) = Entropy( S ) - 
vValues( A ) S
Entropy( S v )

E(S)=0.993
29+, 35- A1 E(S)=0.993 29+, 35- A2
t f t f
0.650 0.522 0.989 0.997
25+, 5- 4+, 30- 15+, 19- 14+, 16-
Gain: 0.395 Gain: 0.000

E(S)=0.993 E(S)=0.993
29+, 35- A1 29+, 35- A2
t f t f
0.708 0.742 0.937 0.619
21+, 5- 8+, 30- 18+, 33- 11+, 2-
Gain: 0.265 Gain: 0.121 15
 Gain function
 Gain is measure of how much can
– Reduce uncertainty
 Value lies between 0,1
 What is significance of
 gain of 0?
 example where have 50/50 split of +/- both before and after
discriminating on attributes values
 gain of 1?
 Example of going from “perfect uncertainty” to perfect certainty
after splitting example with predictive attribute

– Find “patterns” in TE’s relating to attribute values

 Move to locally minimal representation of TE’s

16
Training Examples

Day Outlook Temp Humidity Wind Tennis?

D1 Sunny Hot High Weak No
D2 Sunny Hot High Strong No
D3 Overcast Hot High Weak Yes
D4 Rain Mild High Weak Yes
D5 Rain Cool Normal Weak Yes
D6 Rain Cool Normal Strong No
D7 Overcast Cool Normal Strong Yes
D8 Sunny Mild High Weak No
D9 Sunny Cool Normal Weak Yes
D10 Rain Mild Normal Weak Yes
D11 Sunny Mild Normal Strong Yes
D12 Overcast Mild High Strong Yes
D13 Overcast Hot Normal Weak Yes
D14 Rain Mild High Strong No
Determine the Root Attribute
9+, 5- E=0.940 9+, 5- E=0.940

Humidity Wind

High Low Weak Strong

3+, 4- 6+, 1- 6+, 2- 3+, 3-

E=0.985 E=0.592 E=0.811 E=1.000

Gain (S, Humidity) = 0.151 Gain (S, Wind) = 0.048

Gain (S, Outlook) = 0.246 Gain (S, Temp) = 0.029

18
 Sort the Training Examples
9+, 5- {D1,…,D14}

Outlook

Sunny Overcast Rain

{D1,D2,D8,D9,D11} {D3,D7,D12,D13} {D4,D5,D6,D10,D15}

2+, 3- 4+, 0- 3+, 2-

? Yes ?

Ssunny= {D1,D2,D8,D9,D11}
Gain (Ssunny, Humidity) = .970
Gain (Ssunny, Temp) = .570
Gain (Ssunny, Wind) = .019 19
Final Decision Tree for Example

Outlook

Sunny Rain
Overcast

Humidity
Yes Wind
High
Normal Strong Weak

No Yes No
Yes

20
 Hypothesis Space Search by ID3
• Hypothesis space (all possible trees) is complete!
– Target function is included in there

21
 Hypothesis Space Search in Decision Trees
• Conduct a search of the space of decision trees which
can represent all possible discrete functions.

• Goal: to find the best decision tree

• Finding a minimal decision tree consistent with a set of data

is NP-hard.

• Perform a greedy heuristic search: hill climbing without

backtracking

• Statistics-based decisions using all data

22
Hypothesis Space Search by ID3
• Hypothesis space is complete!
– H is space of all finite DT’s (all discrete functions)
– Target function is included in there
• Simple to complex hill-climbing search of H
– Use of gain as hill-climbing function
• Outputs a single hypothesis (which one?)
– Cannot assess all hypotheses consistent with D (usually many)
– Analogy to breadth first search
 Examines all trees of given depth and chooses best…
• No backtracking
– Locally optimal ...
• Statistics-based search choices
– Use all TE’s at each step
– Robust to noisy data

23
Restriction bias vs. Preference bias
• Restriction bias (or Language bias)
– Incomplete hypothesis space
• Preference (or search) bias
– Incomplete search strategy
• Candidate Elimination has restriction bias
• ID3 has preference bias
• In most cases, we have both a restriction and a
preference bias.

24
Inductive Bias in ID3

• Preference for short trees, and for those with high

information gain attributes near the root
• Principle of Occam's razor
– prefer the shortest hypothesis that fits the data
• Justification
– Smaller likelihood of a short hypothesis fitting the data
at random
• Problems
– Other ways to reduce random fits to data
– Size of hypothesis based on the data representation
 Minimum description length principle

25
Overfitting the Data
• Learning a tree that classifies the training data perfectly may
not lead to the tree with the best generalization performance.
- There may be noise in the training data the tree is fitting
- The algorithm might be making decisions based on
very little data
• A hypothesis h is said to overfit the training data if the is
another hypothesis, h’, such that h has smaller error than h’
on the training data but h has larger error on the test data than h’.

On training

accuracy On testing

Complexity of tree 26
 Overfitting in Decision Trees
• Consider adding noisy training example (should be +):
Day Outlook Temp Humidity Wind Tennis?
D15 Sunny Hot Normal Strong No

• What effect on earlier tree?

Outlook

Sunny Overcast Rain

Humidity Yes Wind

High Normal Strong Weak

27
 Overfitting - Example

Noise or other Outlook

coincidental regularities

Sunny Overcast Rain

1,2,8,9,11 3,7,12,13 4,5,6,10,14
2+,3- 4+,0- 3+,2-
Humidity Yes Wind

High Normal Strong Weak

No Wind No Yes

Strong Weak
No Yes 28
 Avoiding Overfitting

• Two basic approaches

- Prepruning: Stop growing the tree at some point during
construction when it is determined that there is not enough
data to make reliable choices.
- Postpruning: Grow the full tree and then remove nodes
that seem not to have sufficient evidence. (more popular)
• Methods for evaluating subtrees to prune:
- Cross-validation: Reserve hold-out set to evaluate utility (more popular)
- Statistical testing: Test if the observed regularity can be
dismissed as likely to be occur by chance
- Minimum Description Length: Is the additional complexity of
the hypothesis smaller than remembering the exceptions ?
This is related to the notion of regularization that we will see
in other contexts– keep the hypothesis simple.

29
Reduced-Error Pruning
• A post-pruning, cross validation approach
- Partition training data into “grow” set and “validation” set.
- Build a complete tree for the “grow” data
- Until accuracy on validation set decreases, do:
For each non-leaf node in the tree
Temporarily prune the tree below; replace it by majority vote.
Test the accuracy of the hypothesis on the validation set
Permanently prune the node with the greatest increase
in accuracy on the validation test.
• Problem: Uses less data to construct the tree
• Sometimes done at the rules level

General Strategy: Overfit and Simplify

30
Rule post-pruning

• Allow tree to grow until best fit (allow overfitting)

• Convert tree to equivalent set of rules
– One rule per leaf node
– Prune each rule independently of others
 Remove various preconditions to improve
performance
– Sort final rules into desired sequence for use

31
Example of rule post pruning
• IF (Outlook = Sunny) ^ (Humidity = High)
– THEN PlayTennis = No
• IF (Outlook = Sunny) ^ (Humidity = Normal)
– THEN PlayTennis = Yes

Outlook

Sunny Overcast Rain

Humidity Yes Wind

High Normal Strong Weak

32
Extensions of basic algorithm

• Continuous valued attributes

• Attributes with many values
• TE’s with missing data
• Attributes with associated costs
• Other impurity measures
• Regression tree

33
 Continuous Valued Attributes
• Create a discrete attribute from continuous variables
– E.g., define critical Temperature = 82.5
• Candidate thresholds
– chosen by gain function
– can have more than one threshold
– typically where values change quickly
(48+60)/2 (80+90)/2

Temp 40 48 60 72 80 90
Tennis? N N Y Y Y N

34
Attributes with Many Values
• Problem:
– If attribute has many values, Gain will select it (why?)
– E.g. of birthdates attribute
 365 possible values

 Likely to discriminate well on small sample

– For sample of fixed size n, and attribute with N values, as N ->

infinity
 ni/N -> 0

 - pi*log pi -> 0 for all i and entropy -> 0

 Hence gain approaches max value

35
 Attributes with many values
• Problem: Gain will select attribute with many values
• One approach: use GainRatio instead

Gain( S , A)
GainRatio ( S , A) =
SplitInformation( S , A) Entropy of the
partitioning
c Si Si
SplitInformation( S , A) = - log 2 Penalizes
higher number
i =1 S S of partitions

where Si is the subset of S for which A has value vi

(example of Si/S = 1/N: SplitInformation = log N)
36
 Unknown Attribute Values

• What if some examples are missing values of attribute A?

• Use training example anyway, sort through tree
– if node n tests A, assign most common value of A among other
examples sorted to node n
– assign most common value of A among other examples with same
target value
– assign probability pi to each possible value vi of A
 assign fraction pi of example to each descendant in tree
• Classify test instances with missing values in same fashion
• Used in C4.5

37
Attributes with Costs
• Consider
– medical diagnosis: BloodTest has cost $150, Pulse has a cost of $5.
– robotics, Width-From-1ft has cost 23 sec., from 2 ft 10s.
• How to learn a consistent tree with low expected cost?
• Replace gain by
– Tan and Schlimmer (1990)
Gain 2 ( S , A)
Cost ( A)
– Nunez (1988)
2Gain( S , A) - 1
(Cost ( A) + 1)w
where w  [0, 1] determines importance of cost 38
Gini Index
• Another sensible measure of impurity
(i and j are classes)

• After applying attribute A, the resulting Gini index is

• Gini can be interpreted as expected error rate

 Gini Index

. .

. .
. .

Attributes: color, border, dot

Classification: triangle, square

40
 . .
. .
. .
. .
red
Gini Index for Color

Color? green

.
yellow .

.
.

41
Gain of Gini Index

42
Three Impurity Measures

A Gain(A) GainRatio(A) GiniGain(A)

Color 0.247 0.156 0.058
Outline 0.152 0.152 0.046
Dot 0.048 0.049 0.015

43
Decision Trees as Features
• Rather than using decision trees to represent the target function, use
small decision trees as features

• When learning over a large number of features, learning decision trees

is difficult and the resulting tree may be very large
 (over fitting)
• Instead, learn small decision trees, with limited depth.
• Treat them as “experts”; they are correct, but only on a small region in
the domain.
• Then, learn another function over these as features.

44
Regression Tree
• Similar to classification
• Use a set of attributes to predict the value (instead
of a class label)
• Instead of computing information gain, compute
the sum of squared errors
• Partition the attribute space into a set of
rectangular subspaces, each with its own predictor
– The simplest predictor is a constant value

45
 Rectilinear Division
• A regression tree is a piecewise constant function of the
input attributes
X2
X1 t1
r5
r2
X2  t2 X1  t3
r3
t2 r4
r1
r1 r2 r3 X2  t4

t1 t3 X1
r4 r5

46
 Growing Regression Trees

• To minimize the square error on the learning sample,

the prediction at a leaf is the average output of the
learning cases reaching that leaf
• Impurity of a sample is defined by the variance of the
output in that sample:
I(LS)=vary|LS{y}=Ey|LS{(y-Ey|LS{y})2}

• The best split is the one that reduces the most variance:

DI ( LS , A) = vary|LS { y} - 
| LS a |
vary|LS a { y}
a | LS |

47
Regression Tree Pruning
• Exactly the same algorithms apply: pre-pruning
and post-pruning.
• In post-pruning, the tree that minimizes the
squared error on VS is selected.
• In practice, pruning is more important in
regression because full trees are much more
complex (often all objects have a different output
values and hence the full tree has as many leaves
as there are objects in the learning sample)

48
When Are Decision Trees Useful ?
• Advantages
– Very fast: can handle very large datasets with many
attributes
– Flexible: several attribute types, classification and
regression problems, missing values…
– Interpretability: provide rules and attribute importance
• Disadvantages
– Instability of the trees (high variance)
– Not always competitive with other algorithms in terms
of accuracy

49
 History of Decision Tree Research
• Hunt and colleagues in Psychology used full search decision
trees methods to model human concept learning in the 60’s

• Quinlan developed ID3, with the information gain heuristics

in the late 70’s to learn expert systems from examples

• Breiman, Friedmans and colleagues in statistics developed

CART (classification and regression trees simultaneously

• A variety of improvements in the 80’s: coping with noise,

continuous attributes, missing data, non-axis parallel etc.

• Quinlan’s updated algorithm, C4.5 (1993) is commonly used (New:C5)

• Boosting (or Bagging) over DTs is a good general purpose algorithm

50
Summary
• Decision trees are practical for concept learning
• Basic information measure and gain function for best first
search of space of DTs
• ID3 procedure
– search space is complete
– Preference for shorter trees
• Overfitting is an important issue with various solutions
• Many variations and extensions possible

51
 References
• Classification and regression trees, L.Breiman et al.,
Wadsworth, 1984
• C4.5: programs for machine learning, J.R.Quinlan,
Morgan Kaufmann, 1993
• Random Forests, L. Breiman, Leo, Machine Learning
45 (1): 5–32, 2001.
• The elements of statistical learning : Data mining,
inference, and prediction, Hastie, T., Tibshirani, R.,
Friedman, New York: Springer Verlag, 2001.
• Constructing Optimal Binary Decision Trees is NP-
complete, Hyafil, Laurent; Rivest, RL. Information
Processing Letters 5 (1): 15–17, 1976 52
Software
• In R:
– Packages tree and rpart
• C4.5:
– http://www.cse.unwe.edu.au/~quinlan
• Weka
– http://www.cs.waikato.ac.nz/ml/weka

53