Introduction to Machine Learning

Thanh-Nghi Do
Can Tho University
dtnghi@cit.ctu.edu.vn

Can Tho
Dec. 2019
Content

■ Motivation
■ Definitions and problems

■ Applications

2
Content

■ Motivation
■ Definitions and problems

■ Applications

3
 ■ Motivation
■ Definitions & problems
Applications
Motivation
■

■ Why machine learning is important?

● Growing volumes and varieties of available data, computational
processing that is cheaper and more powerful, and affordable
data storage.
● Physical, biological, financial phenomenon, etc.: too complex to
be described in a deterministic way
● Machine learning techniques
● Computers can learn without being programmed to perform
specific tasks
● Voluntary to replace human experts

4
 ■ Motivation
■ Definitions & problems
Applications
Motivation
■

■ Why machine learning is important?

● Adapt to the environment
● Deal with new problems
● Data analysis
● Pattern recognition
● Computer vision
● Control robots, self-driving cars
● Text mining, chatbots
● Information retrieval
● Game, etc

5
Content

■ Motivation
■ Definitions and problems

■ Applications

6
 ■ Motivation
■ Definitions & problems
Applications
Definitions
■

■ Wikipedia
● Learning is the process of acquiring new, or modifying
existing, knowledge, behaviors, skills, values, or preferences

■ Arthur L. Samuel, AI pioneer, 1959

● Machine learning is the field of study that gives computers
the ability to learn without being explicitly programmed

■ Tom Mitchell, Professor at Carnegie Mellon University

● 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 7
 ■ Motivation
■ Definitions & problems
Applications
Definitions
■

8
 ■ Motivation
■ Definitions & problems
Applications
Definitions
■

9
Sebastian Raschka STAT 479: Machine Learning FS 2018
 Figure
Figure 1: Machine learning 1: Machine
vs. ”classic” learning
programming.
■ Motivation
vs. ”class
■ Definitions & problems
Applications
Definitions
■

ncrete, Tom Mitchell’s quote from his Machine Learning book2 :

bit more concrete, Tom Mitchell’s quote from his Ma
“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.”
— “A
Tom computer program
Mitchell, Professor at CarnegieisMellon
saidUniversity
to learn
some class of tasks T and performance
, consider a handwritingHandwriting
in T , asRecognition
recognition learning
measured Example:
problem (from
by P ,Mitchell’s
improves book):with

recognizing and classifying handwritten words within — Tom Mitchell, P

images
ance measure P : percent of words correctly classified
experience E: a database of handwritten words with given classifications
an example, consider a handwriting recognition learn
cations of Machine Learning 10
• Task T : recognizing and classifying handwritten w
 ■ Motivation
■ Definitions & problems
Applications
Machine Learning, AI, Deep Learning
■

11
 ■ Motivation
■ Definitions & problems
Applications
Problems
■

■ Supervised learning
● Given a set of m training examples
DTrain = {(X1, Y1) , . . . ,(Xm, Ym)} such that
Xi Î Rn is the feature vector of the i-th example and
Yi is its label (i.e., target)

● A learning algorithm tries to find a function f: X à Y

Y = f(X) + ε

● Methods: decision trees, neural networks, support vector

machines, k nearest neighbors, naive Bayes, etc
12
 ■ Motivation
■ Definitions & problems
Applications
Classification problem
■

13
 ■ Motivation
■ Definitions & problems
Applications
Classification problem
■

14
 ■ Motivation
■ Definitions & problems
Applications
Classification problem
■

new

15
 ■ Motivation
■ Definitions & problems
Applications
Regression problem
■

16
 ■ Motivation
■ Definitions & problems
Applications
Regression problem
■

17
 ■ Motivation
■ Definitions & problems
Applications
Regression problem
■

new

18
 ■ Motivation
■ Definitions & problems
Applications
Problems
■

■ Unsupervised learning
● Given a set of m training examples without labels
DTrain = {X1, . . . , Xm} such that
Xi Î Rn is the feature vector of the i-th example

● A learning algorithm tries to find the hidden structure (i.e.

clusters) in unlabeled data according to similarities, patterns
and differences without any prior training of data

● Methods: hierarchical clustering, k-means, etc

19
 ■ Motivation
■ Definitions & problems
Applications
Unsupervised learning: clustering
■

20
 ■ Motivation
■ Definitions & problems
Applications
Unsupervised learning: clustering
■

21
 ■ Motivation
■ Definitions & problems
Applications
Challenges
■

■ Statistical techniques
● m, n are reasonable; under assumption of the linear model
and the data distribution is known (Gaussian, Binomial,
Poisson)
■ It seems to be very difficult
● Small samples
● Unknown data distribution
● Unknown target function
● Non-linear models
● Very large data

22
 ■ Motivation
■ Definitions & problems
Applications
Challenges
■

■ Method selection
● Many learning algorithms
● No free lunch in machine learning
● Given a problem => model selection
● Empirical test, tuning hyper-parameters, comparison
● Training time, testing time, model quality

23
Content

■ Motivation
■ Definitions and problems

■ Applications

24
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

■ Bioinformatics
■ Gene expression classification
■ Diagnostic support system
■ Computer vision
■ Handwriting character recognition
■ Speech recognition
■ Fraud detection
■ Financial market analysis

25
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

■ Information retrieval
■ Analyzing and indexing images and vidéos
■ Content-based image retrieval
■ Games
■ Natural language processing
■ Recommendation system
■ Sentiments analysis
■ Robotics
■ Self-driving cars
■ Cyber-security
■ etc.
26
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

27
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

28
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

29
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

30
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

31
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

32
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

33
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

Category: Comp.mac.

34
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

35
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

36
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

37
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

38
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

39
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

40
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

Cancer?

41
 ■ Motivation
■ Definitions & problems
Applications
Applications
■

42