Class 1 X
Class 1 X
Machine Learning
4771
Topic 1
Introduction
Machine Learning: What, Why and Applications
Syllabus, policies, texts, web page
Historical Perspective
Machine Learning Tasks and Tools
Digit Recognition Example
Machine Learning Approach
Deterministic or Probabilistic Approach
Why Probabilistic?
Tony Jebara, Columbia University
About me
Tony Jebara, Associate Professor of Computer Science
Started at Columbia in 2002
PhD from MIT in Machine Learning
Thesis: Discriminative, Generative and Imitative Learning (2001)
www.cs.columbia.edu/learning
Tony Jebara, Columbia University
Syllabus
www.cs.columbia.edu/~jebara/4771/MLInfo.htm
Supervised
x
x xx x x x
x xx
xx x x x
x O x x
OO x x
O O OO x x
O O x x
O
Unsupervised
Feature Selection Detection p(x)<t
x
xxx xx x xx x x x
xxx x x x
x x xx x
x x x
xx x x xxxx
Tony Jebara, Columbia University
0 1 1 1 1
Want to automate zipcode reading in post office
Look at an image and say if it is a 1 or 0
8x8 pixels of gray-level (0.0=dark, 0.5=gray, 1.0=white)
Learn from above labeled training images
Predict labels on testing images ? ? ? ? ?
Binary Classification [0,1]
What to do?
? ? ? ? ?
Tony Jebara, Columbia University
0 1 1 1 1 0.1
0.5
Model Representation
x 0.98
x xx x x x
x 0.5
x x x
x O
OOO OO
O
O O
O
i =1 i Xi + 0 {( )}
D
)( ) (
X1,Y1 , X 2,Y2 ,, X N ,YN
Criterion & Algorithm
x
x xx x x x
x
x x x
( )
x O
y = f (X ) = sign
D
O
O O OO Xi + 0
O i =1 i
O O
O
Tony Jebara, Columbia University