5864281 (1)

INTRODUCTION TO
MACHINE LEARNING
Methods, applications, etc.
Overview
 What is machine learning?
 Applications of machine learning
 Types of algorithms
 Basics of statistical pattern recognition
 Preprocessing
 Feature Extraction/Selection
 Learning
 Final remarks
Overview
 Preprocessing
 Learning
 Final remarks
What is Machine Learning?
 “Field of study that gives computers the ability to
learn without being explicitly programmed”
 Arthur Samuel (1959)
 “A computer program is said to learn from
experience E with respect to some class of tasks T
and performance measure P, if its performance at
tasks in T, as measured by P, improves with
experience E”
 Tom M. Mitchell (1998)
Example: Spam Email Detection
 “A computer program is said to learn from
experience E with respect to some class of tasks T
and performance measure P, if its performance at
tasks in T, as measured by P, improves with
experience E”
 In our project,
 T: classify emails as spam or not spam
 E: watch the user label emails as spam or not spam
 P: percentage of correctly-identified emails
Overview
 Preprocessing
 Learning
 Final remarks
Applications of Machine Learning
 Facial recognition
 Self-customizing programs (Netflix, Amazon, etc.)
 Speech recognition
 Other applications:
 Spam detection
 Internet search
 Data mining medical records
 Many more!
Overview
 Preprocessing
 Learning
 Final remarks
Supervised Learning
 Given labeled training examples
 Find correct prediction for an unlabeled example
 E.g. Handwriting recognition
Types of Supervised Learning Problems
Classification Regression
 Predict whether a  Given the input,

given input belongs to predict a continuous
one of multiple classes output
 Output is discrete  Output is continuous
Unsupervised Learning
 Given unlabeled training samples
 Discover low dimensional patterns, structure
 E.g. Cluster news articles by topic
Reinforcement Learning
 Learn from delayed feedback
 E.g. Machine learns how to play chess
Overview
 Preprocessing
 Learning
 Final remarks
Model of Statistical Pattern Recognition
 Training (learning): classifier is trained to partition the

feature space
 Classification (testing): classifier assigns input pattern
to one of pattern classes based on measured features
Figure: Model for statistical pattern recognition

 Preprocessing: clean up the data

 Separate data from background
 Remove noise

 Feature Extraction/ Selection
 Find the appropriate features for representing the
training examples
 Translate data into a vector space
 Vector has d feature values x = x1 x2 ... xd

Feature Extraction/ Selection
 Feature vector: representation of real world objects
 Choice of representation strongly influences results
 Curse of Dimensionality
 The number of examples needed to train increases
rapidly with the number of features used
 Feature Selection: select a subset of features that
leads to smallest classification errors
 Feature Extraction: create new features based on
transformations or combinations of original features
 Learning
 Theclassifier is trained to identify the correct classes
 Can use various methods to train the classifier

Learning
 Class-conditional densities: probability of feature
vector x belonging to class wi defined as P(x | wi )
 Known:optimal Bayes decision rule can be used
 Unknown: must be learned

Learning
 Form of class-conditional densities
 Parametric: form is known, some parameters unknown
 E.g. we know the data is Gaussian
 Nonparametric: form is unknown
 Either
estimate density function or construct decision
boundary rules

Learning
Figure: various approaches in statistical pattern recognition

Overview
 Preprocessing
 Learning
 Final remarks
Summary
 Machine learning: computers learn from experience
to improve performance of certain tasks
 Many useful applications that we use today
 Different steps to machine learning process
 Preprocessing
 Learning
 Type of problem will affect which approach you

decide to use
Sources
 “Statistical Pattern Recognition: A Review”
 Jain, Anil. K; Duin, Robert. P.W.; Mao, Jianchang (2000).
“Statistical pattern recognition: a review”. IEEE Transtactions
on Pattern Analysis and Machine Intelligence 22 (1): 4-37
 “Machine Learning” Online Course
 Andrew Ng
 http://openclassroom.stanford.edu/MainFolder/CoursePage
.php?course=MachineLearning
 “Machine Learning” Course
 Kilian Weinberger
 http://www.cse.wustl.edu/~kilian/cse517a2010/
Image Sources
 http://googlenewsblog.blogspot.com/2009_05_01_ar
chive.html
 https://www.coursera.org/course/ml
 http://visionwang.com/2009/01/11/apples-iphoto-vs-
googles-picasa-technology-and-strategy/
 http://www.techhive.com/article/2024221/
 http://forums.macrumors.com/showthread.php?t=4804
28
 http://www.stanford.edu/class/cs224s/
 http://en.wikipedia.org/wiki/Machine_learning
Thank you
Any questions?

5864281 (1)

Uploaded by

Copyright:

Available Formats

5864281 (1)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5864281 (1)

Uploaded by

Copyright:

Available Formats

INTRODUCTION TO

 Data mining medical records

 Predict whether a  Given the input,

 Training (learning): classifier is trained to partition the

Figure: Model for statistical pattern recognition

 Preprocessing: clean up the data

Figure: Model for statistical pattern recognition

Figure: Model for statistical pattern recognition

Figure: Model for statistical pattern recognition

Figure: Model for statistical pattern recognition

Figure: Model for statistical pattern recognition

Figure: various approaches in statistical pattern recognition

 Type of problem will affect which approach you

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.