Machine Learning and Data Mining: 02 Machine Learning


Published on

Course "Machine Learning and Data Mining" for the degree of Computer Engineering at the Politecnico di Milano. This lecture gives a very short introduction to the three main machine learning paradigms.

Published in: Technology, Education
1 Comment
No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Machine Learning and Data Mining: 02 Machine Learning

  1. 1. .‘i ¢'! |!NEl‘: °fi‘Il'! 'l! Ell MIL‘-. :I‘| !‘ Machine Learning Machine Learning and Data Mining
  2. 2. Lecture outline 2 Cl What is Machine Learning? El What are the paradigm? i- Unsupervised Learning c- Supervised Learning c: - Reinforcement Learning Proi. Pier Luca Lanzi L noun: -mi-io~ pll 5.i", [il‘“! ‘
  3. 3. What is Machine Learning? 3 Cl “The field of machine learning is concerned with the question of how to construct computer programs that automatically improve with experience. ” Tom Mitchell (1997) El A 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. El A well-defined learning task is defined by P, T, and E. Prof. Pier Luca Lanzi i 7"| !|l,1'1'. ‘| h-lo -in . '.IIl. ;wu
  4. 4. Example: checkers 4 Cl Task T: playing checkers CI Artificial Intelligence c- Design and implement a computer-based system that exhibit intelligent action El Machine Learning : .. Write a program that can learn how to play : - It can learn from examples of previous games, by playing against another opponent, by playing against itself Prol. Pier Luca Lanzi : IoI, I|II, :I-ml-iol pll | .l"_L| l‘“! ‘
  5. 5. Examples 5 Cl A checker learning problem : - Task T: playing checkers : - Performance P: percent of games won against opponents c- Training experience E: playing practice games E! A handwriting recognition learning problem t- Task T: recognizing and classifying handwritten words withing images i—- Performance P: percent of words correctly classified : — Training experience E: a database of handwritten words with given classification Prof. Pier Luca Lanzi tlvllllil-£. ‘llv-lo‘ pll . '.I| l.; .‘|9*
  6. 6. Example 5 Cl A robot driving learning problem : - Task T: driving on public four-lane highways using vision l- Performance P: average distance traveled before an error c- Training experience E: a sequence of images and steering commands recorded while observing a human driver Prof. Pier Luca Lanzi K tlvlllla-lellwiol pll . '.| |l. ;w|9:
  7. 7. I . L. 4! l. ‘ C. ‘ lb - _- «L ‘ 1 J, [’ § 5 iv '1 Prol. Pier Luca Lanzi >X? '!| ,I| ;.1'l. ‘Il-IE0‘ Pll . '.Illi/ :h‘l9l
  8. 8. What is unsupervised learning? 3 Cl Task T: finding interesting groups into data, learning “what normally happens” [3 Performance P: how good, how interesting the groups are Cl Training experience E: raw data El Example applications : - Customer segmentation in CRM : — Color quantization for image compression, :» Bioinformatics Prol. Pier Luca Lanzi : 'oI, I|II, :t-l. ‘ll-iol pll u. l"_L| l‘“! ‘
  9. 9. What is an apple? 9 wf’ l1 s -" 1. ! ’ = r:fl; .¥‘‘{m r s ‘try. ’ S, -‘: .‘| ' ‘Z’ ‘ R . o . ," . -' ‘it’ I u 9;» L 5’ Prol. Pier Luca Lanzi D/ t'! |,IN,1'! .‘Il-K0‘ Pll . '.Illi: wl9l
  10. 10. What is an apple? 10 Prol. Pier Luca Lanzi I : I9I_ul: -llllctm pll . '.Illi: It‘I9=7
  11. 11. Are these apples? 11 it 2; " ‘N I - . . I f r“. . , .. _ __ 7 , £ Prol. Pier Luca Lanzi i >1 : I9I_| Il: -llllctol pll . '.Illi: wl9l 7
  12. 12. What is supervised learning? 12 Cl Training experience E: examples labeled by a supervisor CI Task T: to extract a description of a concept from the data. Use the description to predict the output for future examples Cl Performance P: how accurate the description is El Example applications —- Credit approval - Target marketing — Medical diagnosis . - Fraud detection -—s rv rv r—v Prol. Pier Luca Lanzi : 'oI, I|II, :t-l. ‘ll-iol pll . '.Ill1;h‘l9:
  13. 13. What is Reinforcement Learning? 13 Prof. Pier Luca Lanzi : I9I| u:vl. ‘lnt9« pll . '.m; m91
  14. 14. What is Reinforcement Learning? 14 H Agent I Environment - CI The agent learn through trial-and-error interactions CI The goal is to maximize the amount of reward received from the environment El Compute a value function Q(s, ,a ) mapping state-action pairs into expected future payoffs Prol. Pier Luca Lanzi — POLIYECNICO DI MILANO
  15. 15. What is reinforcement learning? 15 Cl Training experience E: online interactions with the environment Ci Task T: collect as much reward as possible Cl Performance P: the amount of reward El Example applications 1- Robot learning : — Games : » Multiagent learning Prol. Pier Luca Lanzi : 'oI, I|II, :t-l. ‘ll-iol pll . '.Ill1;h‘l9:
  16. 16. Algorithms, Paradigms, Applications C]Applications : . Agents : » Data Mining ‘ : » Robotics ElAlgorithms 1- Clustering I I I E 16 DParadigms » Unsupervised Learning - Supervised Learning - Reinforcement Learning r» Association Rules 1» Decision trees it . . Prol. Pier Luca Lanzi T'9II",1‘C‘"¢I9‘ Pll 1'1"}: :t‘I9‘
  17. 17. Machine Learning and Data Mining 17 Cl Machine learning algorithms acquire structural descriptions from examples El Structural descriptions represent patterns explicitly 1-. They can be used to predict outcome in new situations r- They can be used to understand and explain how prediction is derived El Unsupervised learning : - Clustering : - Association rules Cl Supervised learning :1 Decision trees :1 Decision rules : » Bayesian classifiers Prol. Pier Luca Lanzi : 'oI, I|II; t-calla-iol pll . '.Ill1;h‘l9: