SLearning

Information about SLearning

Published on November 30, 2007

Author: Felipe

Source: authorstream.com

Content

Supervised Learning:  Supervised Learning Introduction to Artificial Intelligence COS302 Michael L. Littman Fall 2001 Administration:  Administration Exams graded! http://www.cs.princeton.edu/courses/archive/fall01/cs302/whats-new.html Project groups. Supervised Learning:  Supervised Learning Most studied in machine learning. http://www1.ics.uci.edu/~mlearn/MLRepository.html Set of examples (usually numeric vectors). Split into: Training: Allowed to see it Test: Want to minimize error here Another Significant App:  Another Significant App Name A B C D E F G 1. Jeffrey B. 1 0 1 0 1 0 1 - 2. Paul S. 0 1 1 0 0 0 1 - 3. Daniel C. 0 0 1 0 0 0 0 - 4. Gregory P. 1 0 1 0 1 0 0 - 5. Michael N. 0 0 1 1 0 0 0 - 6. Corinne N. 1 1 1 0 1 0 1 + 7. Mariyam M. 0 1 0 1 0 0 1 + 8. Stephany D. 1 1 1 1 1 1 1 + 9. Mary D. 1 1 1 1 1 1 1 + 10. Jamie F. 1 1 1 0 0 1 1 + Features:  Features A: First name ends in a vowel? B: Neat handwriting? (Lisa test.) C: Middle name listed? D: Senior? E: Got extra-extra credit? F: Google brings up home page? G: Google brings up reference? Decision Tree:  Decision Tree Internal nodes: features Leaves: classification F A D A 0 1 8,9 2,3,7 1,4,5,6 10 Error: 30% Search:  Search Given a set of training data, pick a decision tree: search problem! Challenges: Scoring function? Large space of trees. Scoring Function:  Scoring Function What’s a good tree? Low error on training data Small Small tree is obviously not enough, why isn’t low error? Low Error Not Enough:  Low Error Not Enough C E B 0 1 F middle name? EEC? Neat? Google? Training set Error: 0% (can always do this?) Memorizing the Data:  Memorizing the Data D E F “Learning Curve”:  “Learning Curve” error Tree size What’s the Problem?:  What’s the Problem? Memorization w/o generalization Want a tree big enough to be correct, but not so big that it gets distracted by particulars. But, how can we know? (Weak) theoretical bounds exist. Cross-validation:  Cross-validation Simple, effective hack method. Data Test Train C-V Train’ Concrete Idea: Pruning:  Concrete Idea: Pruning Use Train’ to find tree w/ no error. Use C-V to score prunings of tree. Return pruned tree w/ max score. How Find the Tree?:  How Find the Tree? Lots to choose from. Could use local search. Greedy search… Why Might This Fail?:  Why Might This Fail? No target function, just noise Target function too complex (22^n possibilities, parity) Training data doesn’t match target function (PAC bounds) Theory: PAC Learning:  Theory: PAC Learning Probably Approximately Correct Training/testing from distribution. With probability 1-d, learned rule will have error smaller than e. Bounds on size of training set in terms of d, e, “dimensionality” of the target concept. Classification:  Classification Naïve Bayes classifier Differentiation vs. modeling More on this later. What to Learn:  What to Learn Decision tree representation Memorization problem: causes and cures (cross-validation, pruning) Greedy heuristic for finding small trees with low error Homework 9 (due 12/5):  Homework 9 (due 12/5) Write a program that decides if a pair of words are synonyms using wordnet. I’ll send you the list, you send me the answers. Draw a decision tree that represents (a) f1+f2+…+fn (or), (b) f1f2…fn (and), (c) parity (odd number of features “on”). More soon

Related presentations


Other presentations created by Felipe

Flowering Plants
14. 12. 2007
0 views

Flowering Plants

Cheer and Dance
23. 11. 2007
0 views

Cheer and Dance

discgolfpresentation
05. 11. 2007
0 views

discgolfpresentation

video enhanced learning VC2002
03. 10. 2007
0 views

video enhanced learning VC2002

utah biomes
26. 11. 2007
0 views

utah biomes

LSC 1
05. 12. 2007
0 views

LSC 1

comedyOrTragedy
25. 10. 2007
0 views

comedyOrTragedy

EvenettJenny2
24. 10. 2007
0 views

EvenettJenny2

SBI billings varnell
07. 11. 2007
0 views

SBI billings varnell

CrewBar QuizShow5 Respostas
13. 11. 2007
0 views

CrewBar QuizShow5 Respostas

ZOONOSIS ASSESSMENTS PRECAUTIONS
16. 11. 2007
0 views

ZOONOSIS ASSESSMENTS PRECAUTIONS

orientation7
05. 11. 2007
0 views

orientation7

part4
26. 11. 2007
0 views

part4

Williams e
21. 12. 2007
0 views

Williams e

NC HR JAL Presentation
24. 12. 2007
0 views

NC HR JAL Presentation

Nixon
25. 12. 2007
0 views

Nixon

Flower Power short
02. 01. 2008
0 views

Flower Power short

Robots In Manufacturing
04. 01. 2008
0 views

Robots In Manufacturing

Tick talk Hurt
04. 01. 2008
0 views

Tick talk Hurt

rzeki
09. 10. 2007
0 views

rzeki

dhs16 139539
07. 01. 2008
0 views

dhs16 139539

2b ROM p1
30. 10. 2007
0 views

2b ROM p1

Haas
28. 12. 2007
0 views

Haas

1 purpose definitions
25. 10. 2007
0 views

1 purpose definitions

UBS Hoefert
25. 10. 2007
0 views

UBS Hoefert

von zengen
20. 11. 2007
0 views

von zengen

ECDLuaf
31. 10. 2007
0 views

ECDLuaf

olgariia
20. 02. 2008
0 views

olgariia

Products and Markets
24. 02. 2008
0 views

Products and Markets

Module 08 revised 6 06
27. 02. 2008
0 views

Module 08 revised 6 06

JSExtn
29. 12. 2007
0 views

JSExtn

me5Revolution
29. 02. 2008
0 views

me5Revolution

air4b
05. 03. 2008
0 views

air4b

japan tadteramoto06
27. 03. 2008
0 views

japan tadteramoto06

19th Unicode officexp c
30. 03. 2008
0 views

19th Unicode officexp c

Windemuth
26. 10. 2007
0 views

Windemuth

Hubberts Peak
13. 04. 2008
0 views

Hubberts Peak

WWIIToModernComputing
05. 01. 2008
0 views

WWIIToModernComputing

alcatel 1
29. 11. 2007
0 views

alcatel 1

Chap5
17. 12. 2007
0 views

Chap5

DAdams
30. 10. 2007
0 views

DAdams

CaribPluralism1
28. 11. 2007
0 views

CaribPluralism1

Tarantino
29. 10. 2007
0 views

Tarantino

mbarrel062002
31. 10. 2007
0 views

mbarrel062002

internetmarketing
28. 11. 2007
0 views

internetmarketing

Speaker 10 King CDC
26. 11. 2007
0 views

Speaker 10 King CDC

25 martie Treaty of Rome
30. 10. 2007
0 views

25 martie Treaty of Rome

TT 3000SSA
07. 11. 2007
0 views

TT 3000SSA

poverty 05
28. 11. 2007
0 views

poverty 05

adecbriefing040506
06. 11. 2007
0 views

adecbriefing040506

BioUML SBML
15. 11. 2007
0 views

BioUML SBML

Christy Hansen
31. 12. 2007
0 views

Christy Hansen

boekel
14. 11. 2007
0 views

boekel

cantanti
21. 11. 2007
0 views

cantanti

en pressrelease gm 1175181932
20. 11. 2007
0 views

en pressrelease gm 1175181932

LatinAmerica
22. 11. 2007
0 views

LatinAmerica

V Info Grosser EU Sieg
16. 11. 2007
0 views

V Info Grosser EU Sieg

climatemurrray
24. 10. 2007
0 views

climatemurrray

marcoeuropeo170204
01. 11. 2007
0 views

marcoeuropeo170204