Lecture21 - CS440/ECE448: Intro to Articial Intelligence!...

Info iconThis preview shows pages 1–12. Sign up to view the full content.

View Full Document Right Arrow Icon
Lecture 21: Classifcation; Decision Trees Prof. Julia Hockenmaier juliahmr@illinois.edu http://cs.illinois.edu/fa11/cs440 CS440/ECE448: Intro to ArtiFcial Intelligence
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Supervised learning: classifcation
Background image of page 2
Supervised learning Given a set D of N items x i , each paired with an output value y i = f( x i ) , discover a function h( x ) which approximates f( x ) D = {( x 1 , y 1 ),… ( x N , y N )} Typically, the input values x are (real-valued or boolean) vectors : x i ˥ R n or x i {0,1} n The output values y are either boolean (binary classifcation) , elements of a Fnite set (multiclass classifcation) , or real (regression)
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Supervised learning Training: fnd h( x ) Given a training set D train oF items ( x i , y i = f( x i )) , return a Function h( x ) which approximates f( x ) Testing: how well does h( x ) generalize? Given a test set D test oF items x i that is disjoint From D train , evaluate how close h( x ) is to f( x ) . (classifcation) accuracy: pctg. oF x i ˥ D test : h( x i ) = f( x i ) 4 CS440/ECE448: Intro AI train test
Background image of page 4
N -fold cross-validation A better indication of how well h(x) generalizes: Split data into N equal-sized parts, Run and evaluate N experiments Report average accuracy, variance, etc. 5 CS440/ECE448: Intro AI
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
The Naïve Bayes Classifer Each item has a number of attributes A 1 =a 1 ,…,A n =a n We predict the class c based on c = argmax c i P(A i = a i | C=c) P(C=c) 6 CS440/ECE448: Intro AI C A1 A2 An
Background image of page 6
An example Can you train a Naïve Bayes classifer to predict whether the customer wants sugar or not? What is P(coFFee | sugar)? 7 CS440/ECE448: Intro AI x1 x2 Y A1: drink A2: milk? C: sugar? coFFee no yes coFFee yes no tea yes yes tea no no
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Questions that came up in class… What are the independence assumptions that Naïve Bayes makes? Are drink and milk independent R.V.s? Are they conditionally independent, given sugar? What happens when your Bayes Net makes independence assumptions that are incorrect? 8 CS440/ECE448: Intro AI
Background image of page 8
Decision trees
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Decision trees In this example, the attributes (drink; milk?) are not conditionally independent given the class ( ʻ sugar ʼ ) 10 CS440/ECE448: Intro AI drink? milk? milk? coffee tea yes no no sugar sugar yes no sugar no sugar
Background image of page 10
Test 2
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 12
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 39

Lecture21 - CS440/ECE448: Intro to Articial Intelligence!...

This preview shows document pages 1 - 12. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online