Machine learning

47
Machine Learning Overview Ricardo Wendell June 2013

description

 

Transcript of Machine learning

Page 1: Machine learning

Machine Learning Overview

Ricardo Wendell

June 2013

Page 2: Machine learning

2

Artificial Intelligence & Machine Learning Artificial Neural Networks Clustering Genetic Algorithms Reinforcement Learning Q&A

Agenda

Page 3: Machine learning

3

Artificial Intelligence & Machine Learning

Page 4: Machine learning

What is Intelligence?

4

Page 5: Machine learning

What’s involved in intelligence?

ü Ability to interact with the real world

ü Reasoning and planning

ü Learning and adaptation

5

Page 6: Machine learning

What’s involved in intelligence?

ü Ability to interact with the real world

ü Reasoning and planning

ü Learning and adaptation

6

Page 7: Machine learning

“To learn is to gain knowledge, or

understanding of, or skill in, by study, instruction, or experience”

7

Page 8: Machine learning

Machine Learning

Field of study that gives computers the ability to learn without being explicitly programmed

8

Page 9: Machine learning

Why?

9

Page 10: Machine learning

Why?

Why should machines have to learn?

10

Page 11: Machine learning

Why?

Why should machines have to learn? Why not design machines to perform as desired in the first place?

11

Page 12: Machine learning

Some reasons…

There are tasks that cannot be defined well except by example

It is possible that hidden among large piles of data are important relationships and correlations

12

Page 13: Machine learning

An example

13 http://googleresearch.blogspot.com.br/2012/08/speech-recognition-and-deep-learning.html

Google’s Deep Learning

Page 14: Machine learning

How can machines learn?

14

Page 15: Machine learning

Types of learning

Supervised

Page 16: Machine learning

Types of learning

Unsupervised

16

Supervised

Page 17: Machine learning

Ways of address these problems…

ü Statistics

ü Brain Models

ü Adaptive Control

ü Evolutionary Models

ü Psychological Models

17

Page 18: Machine learning

Ways of address these problems…

ü Statistics

ü Brain Models

ü Adaptive Control

ü Evolutionary Models

ü Psychological Models

18

Page 19: Machine learning

19

Artificial Neural Networks

Page 20: Machine learning

Biological Inspiration

20

Page 21: Machine learning

McCulloch-Pitts Neuron (1943)

21

Page 22: Machine learning

Multilayer Perceptron

22

Page 23: Machine learning

How do we train it?

23

Page 24: Machine learning

Backpropagation Algorithm initialize  the  weights  in  the  network  

do  

   for  each  example  e  in  the  training  set  

       O  =  neural-­‐net-­‐output(network,  i)  

       T  =  teacher  output  for  e  

       compute  error  (T  -­‐  O)  at  the  output  units  

       compute  delta_wh  for  all  weights  from  hidden  layer  to  output  layer  

       compute  delta_wi  for  all  weights  from  input  layer  to  hidden  layer  

       update  the  weights  in  the  network  

until  all  examples  classified  correctly  or  stopping  criterion  satisfied  

return  the  network  

24

Page 25: Machine learning

Recurrent ANN

25

Page 26: Machine learning

Drawbacks

ANNs with many hidden layers can solve difficult problems… but are very hard to train!

26

Page 27: Machine learning

Deep Learning http://www.kurzweilai.net/how-bio-inspired-deep-learning-keeps-winning-competitions 27

Page 28: Machine learning

28

Clustering

Page 29: Machine learning

Clustering algorithms

29

Page 30: Machine learning

Clustering problems

http://jsfiddle.net/8NpNp/27/ 30

Page 31: Machine learning

Many applications

Medical imaging

Market research

Climatology

Social networks analysis

Recommender systems

31

Page 32: Machine learning

32

Genetic Algorithms

Page 33: Machine learning

Given a population, how can we find the best

group of individuals that satisfy some criteria?

33

Page 34: Machine learning

Given a population, how can we find the best

group of individuals that satisfy some criteria?

34

Page 35: Machine learning

Typical requirements

a genetic representation of the solution domain

a fitness function to evaluate the solution domain

35

Page 36: Machine learning

Genetic representation

Each solution is represented as a “chromosome”

Each solution has a fitness value

36

Page 37: Machine learning

Generic Genetic Algorithm

37

Page 38: Machine learning

Example: Mona Lisa from 1500 characters

38 http://www.youtube.com/watch?v=TManzvC9pi8&NR

Page 39: Machine learning

39

Reinforcement Learning

Page 40: Machine learning

Main concepts

Inspired by behaviorist psychology

Learning by interacting with the environment

40

Suited for problems which include a long-term versus short-term reward trade-off

Page 41: Machine learning

How to teach a computer to play games?

41

Page 42: Machine learning

Markov Decision Processes

42

Page 43: Machine learning

How to flip pancakes?

43

http://www.youtube.com/watch?v=W_gxLKSsSIE

Page 44: Machine learning

44

Conclusion

Page 45: Machine learning

Conclusion There are many techniques…

Results are heavily influenced by input representation

It’s a math-heavy field!

45

Page 46: Machine learning

Some libraries

46

http://www.youtube.com/watch?v=WB9zr0IZCPQ

Page 47: Machine learning

47

Q&A