0% found this document useful (0 votes)
6 views21 pages

Ensemble Learning Techniques Explained

The document discusses ensemble learning techniques in machine learning, highlighting the benefits of combining multiple classifiers to improve prediction accuracy and reduce error rates. It covers methods such as bagging and boosting, explaining how they work and their advantages, including variance reduction and better handling of noise. Additionally, it addresses issues related to ensemble methods, such as parallelism and the independence of classifiers.

Uploaded by

cs21b012
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
6 views21 pages

Ensemble Learning Techniques Explained

The document discusses ensemble learning techniques in machine learning, highlighting the benefits of combining multiple classifiers to improve prediction accuracy and reduce error rates. It covers methods such as bagging and boosting, explaining how they work and their advantages, including variance reduction and better handling of noise. Additionally, it addresses issues related to ensemble methods, such as parallelism and the independence of classifiers.

Uploaded by

cs21b012
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Lecture
Ensemble methods

1
Which technique is better?
 Many techniques
 SVM
 K-nearest neighbours
 Logistic regression
 HMM
 Neural networks

 Which one we should pick?

 Which will perform better?

2
Ensemble Learning
 Each learning technique yields a different hypothesis
 None of these are perfect hypothesis
 Can we combine several imperfect hypothesis to produce a
better hypothesis?
 Hopefully

 Example: Selection of Candidates by voting


 An individual often make mistakes, but majority less likely to make
mistakes
 Example: A committee combines the experts’ opinion and make
a decision
 An individual may have partial knowledge, but when combining the
knowledge of a pool of experts leads to better decision

3
Ensemble Learning
 Ensemble learning  select a collection (ensemble) of
hypotheses and combine their predictions.
 When combing multiple independent and diverse decisions each
of which is at least more accurate than random guessing,
random errors cancel each other out, correct decisions are
reinforced.
 Example:
 generate 50 different decision trees from the same or different
training set and have them vote on the best classification for a new
example.
 Motivation:
 reduce the error rate. Hope is that it will become much more
unlikely that the ensemble of will misclassify an example.

4
Ensemble Learning
 Another way of thinking about ensemble learning:

  way of enlarging the hypothesis space, i.e., the


ensemble itself is a hypothesis and the new hypothesis
space is the set of all possible ensembles constructible
form hypotheses of the original space.
Increasing power of ensemble learning:

Three linear threshold hypothesis


(positive examples on the non-shaded side);
Ensemble classifies as positive any example classified
positively be all three. The resulting triangular region hypothesis
is not expressible in the original hypothesis space.

Linear separator to Polytope


5
Ensemble learning

[Link] 6
Why does it work?
 Intuition
 combining diverse, independent opinions in human decision-
making as a protective mechanism (e.g. stock portfolio)
 Uncorrelated error reduction
 Suppose we have 5 completely independent classifiers for majority
voting
 If accuracy is 70% for each
 10 (.7^3)(.3^2)+5(.7^4)(.3)+(.7^5)
 83.7% majority vote accuracy
 101 such classifiers
 99.9% majority vote accuracy

7
7
Why does it work?

Some unknown distribution

Model 1 Model 6
Model 3 Model 5
Model 2 Model 4

Ensemble gives the global picture! 8


Why does it work?
 Overcome limitations of single hypothesis
 The target function may not be implementable with
individual classifiers, but may be approximated by model
averaging

Decision Tree Model Averaging 9


Learning Ensembles
 Learn multiple alternative definitions of a concept using
different training data or different learning algorithms.
 Combine decisions of multiple definitions, e.g. using
weighted voting.
Training Data

Step 1 : Create
multiple data sets
Data1 Data2  Data m

Learner1 Learner2  Learner m

Step 2 : Build multiple


models Model1 Model2  Model m

Step 3 : Combine Model Combiner Final Model


classifiers
10
Homogenous Ensembles
 Use a single, arbitrary learning algorithm but manipulate
training data to make it learn multiple models.
 Data1  Data2  …  Data m
 Learner1 = Learner2 = … = Learner m
 Different methods for changing training data:
 Bagging: Resample training data
 Boosting: Reweight training data
 DECORATE: Add additional artificial training data

11
Bagging

Ensemble of Hypothesis
Instance
Classification
P1 P2
P4 X Majority {P1(X), P2(X), P3(X), P4(X), P5(X)}
P3 P5

 Classification is wrong only if 3 out of 5 hypothesis is


wrong

12
Parallel Ensemble Learning(Bagging)

 Bagging, is a machine learning ensemble meta-algorithm intended


to improve the strength and accuracy of machine learning
algorithms used in classification and regression purpose. It
additionally diminishes fluctuation of data(variance)and help to
from over-fitting.
 Parallel ensemble methods where the base learners are generated
in parallel
 Algorithms : Random Forest, Bagged Decision Trees, Extra Trees

13
Bagging (bootstrap aggregating)
 Create ensembles by repeatedly randomly resampling the
training data (Brieman, 1996).
 Given a training set of size n, create m samples of size n
by drawing n examples from the original data, with
replacement.
 Combine the m resulting models using simple majority
vote.
 Decreases error by decreasing the variance in the results
due to unstable learners, algorithms (like decision trees)
whose output can change dramatically when the training
data is slightly changed.

14
Advantages of Bagging
 Easy to implement
 Reduces variance - strong beneficial effect on high
variance classifiers
 As the prediction is an average of many classifiers, you
obtain a mean score and variance. Latter can be
interpreted as the uncertainty of the prediction.
 Bagging provides an unbiased estimate of the test error,
(out-of-bag error)

15
Bagging- Final Points
 Works well if the base classifiers are unstable
 Increased accuracy because it reduces variance as
compared to any individual classifier
 Multiple models can be learned in parallel

16
Weighted majority
 However
 In practice, hypothesis may not be independent
 Some hypothesis may perform better than others
 Can take weighted majority
 Decrease weight of correlated hypotheses
 Increase weight of good hypotheses

17
Boosting
 Popular ensemble technique
 Computes a weighted majority
 Can “boost” a “weak learner”
 Operates on a weighted training set

18
Sequential Ensemble learning (Boosting) :
 Boosting, is a machine learning ensemble meta-algorithm
for principally reducing bias, and furthermore variance in
supervised learning, and a group of machine learning
algorithms that convert weak learner to string ones.
 Sequential ensemble methods where the base learners are
generated sequentially.
 Example : Adaboost, Stochastic Gradient Boosting

19
Issues in Ensembles
 Parallelism in Ensembles: Bagging is easily
parallelized, Boosting is not.
 Variants of Boosting to handle noisy data.
 How “weak” should a base-learner for Boosting
be?
 What is the theoretical explanation of boosting’s
ability to improve generalization?
 Exactly how does the diversity of ensembles
affect their generalization performance.
 Combining Boosting and Bagging.

20
Issues in Ensembles
 Can generate multiple classifiers by
 Resampling training data
 Reweighting training data
 Varying learning parameters
 Varying feature set

 Boosting can’t be run in parallel and is very


sensitive to noise, but (in the absence of noise)
seems very insensitive to overfitting
 Ability of ensemble methods to reduce variance
depends on independence of learned classifiers

21
21

You might also like