# 神经网络学习

## 展开查看详情

1. Learning I Linda Shapiro EE/CSE 576 1

2. Learning • AI/Vision systems are complex and may have many parameters. • It is impractical and often impossible to encode all the knowledge a system needs. • Different types of data may require very different parameters. • Instead of trying to hard code all the knowledge, it makes sense to learn it. 2

3. Learning from Observations • Supervised Learning – learn a function from a set of training examples which are preclassified feature vectors. feature vector class (shape,color) Given a previously unseen (square, red) I feature vector, what is the (square, blue) I rule that tells us if it is in (circle, red) II class I or class II? (circle blue) II (triangle, red) I (triangle, green) I (circle, green) ? (ellipse, blue) II (triangle, blue) ? (ellipse, red) II 3

4. Real Observations %Training set of Calenouria and Dorenouria @DATA 0,1,1,0,0,0,0,0,0,1,1,2,3,0,1,2,0,0,0,0,0,0,0,0,1,0,0,1, 0,2,0,0,0,0,1,1,1,0,1,8,0,7,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0, 3,3,4,0,2,1,0,1,1,1,0,0,0,0,1,0,0,1,1,cal 0,1,0,0,0,1,0,0,0,4,1,2 ,2,0,1,0,0,0,0,0,1,0,0,3,0,2,0,0,1,1,0,0,1,0,0,0,1,0,1,6,1,8,2,0,0, 0,0,1,0,0,0,0,0,0,0,0,0,0,1,1,0,0,1,2,0,5,0,0,0,0,0,0,0,1,3,0,0,0,0, 0,cal 0,0,1,0,1,0,0,1,0,1,0,0,1,0,3,0,1,0,0,2,0,0,0,0,1,3,0,0,0,0,0,0,1,0, 2,0,2,0,1,8,0,5,0,1,0,1,0,1,1,0,0,0,0,0,0,0,0,0,2,2,0,0,3,0,0,2,1,1, 5,0,0,0,2,1,3,2,0,1,0,0,cal 0,0,0,0,0,0,0,0,2,0,0,1,2,0,1,1,0,0,0,1 ,0,0,0,0,0,0,0,0,0,1,0,0,0,1,0,0,3,0,0,4,1,8,0,0,0,1,0,0,0,0,0,1,0,1 ,0,1,0,0,0,0,0,0,4,2,0,2,1,1,2,1,1,0,0,0,0,2,0,0,2,2,cal ... 4

5. Learning from Observations • Unsupervised Learning – No classes are given. The idea is to find patterns in the data. This generally involves clustering. • Reinforcement Learning – learn from feedback after a decision is made. 5

6. Topics to Cover • Inductive Learning – decision trees – ensembles – neural nets – kernel machines • Unsupervised Learning – K-Means Clustering – Expectation Maximization (EM) algorithm 6

7. Decision Trees • Theory is well-understood. • Often used in pattern recognition problems. • Has the nice property that you can easily understand the decision rule it has learned. 7

8.Classic ML example: decision tree for “Shall I play tennis today?” 8 from Tom Mitchell’s ML book

9. A Real Decision Tree (WEKA) part23 < 0.5 Dorenouria | part29 < 3.5 | | part34 < 0.5 | | | part8 < 2.5 | | | | part2 < 0.5 | | | | | part63 < 3.5 | | | | | | part20 < 1.5 : dor (53/12) [25/8] Calenouria | | | | | | part20 >= 1.5 | | | | | | | part37 < 2.5 : cal (6/0) [5/2] | | | | | | | part37 >= 2.5 : dor (3/1) [2/0] | | | | | part63 >= 3.5 : dor (14/0) [3/0] | | | | part2 >= 0.5 : cal (21/8) [10/4] | | | part8 >= 2.5 : dor (14/0) [14/0] | | part34 >= 0.5 : cal (38/12) [18/4] | part29 >= 3.5 : dor (32/0) [10/2] part23 >= 0.5 | part29 < 7.5 : cal (66/8) [35/12] | part29 >= 7.5 | | part24 < 5.5 : dor (9/0) [4/0] | | part24 >= 5.5 : cal (4/0) [4/0] 9

10.Evaluation Correctly Classified Instances 281 73.5602 % Incorrectly Classified Instances 101 26.4398 % Kappa statistic 0.4718 Mean absolute error 0.3493 Root mean squared error 0.4545 Relative absolute error 69.973 % Root relative squared error 90.7886 % Total Number of Instances 382 === Detailed Accuracy By Class === TP Rate FP Rate Precision Recall F-Measure ROC Area Class 0.77 0.297 0.713 0.77 0.74 0.747 cal 0.703 0.23 0.761 0.703 0.731 0.747 dor Wg Avg. 0.736 0.263 0.737 0.736 0.735 0.747 === Confusion Matrix === Precision = TP/(TP+FP) Recall = TP/(TP+FN) a b <-- classified as F-Measure = 2 x Precision x Recall 144 43 | a = cal Precision + Recall 58 137 | b = dor 10

11. Properties of Decision Trees • They divide the decision space into axis parallel rectangles and label each rectangle as one of the k classes. • They can represent Boolean functions. • They are variable size and deterministic. • They can represent discrete or continuous parameters. • They can be learned from training data. 11

12.Learning Algorithm for Decision Trees Growtree(S) /* Binary version */ if (y==0 for all (x,y) in S) return newleaf(0) else if (y==1 for all (x,y) in S) return newleaf(1) else choose best attribute xj S0 = (x,y) with xj = 0 S1 = (x,y) with xj = 1 return new node(xj, Growtree(S0), Growtree(S1)) end How do we choose the best attribute? What should that attribute do for us? 12

13. Shall I play tennis today? Which attribute should be selected? “training data” 13 witten&eibe

14. Criterion for attribute selection • Which is the best attribute? – The one that will result in the smallest tree – Heuristic: choose the attribute that produces the “purest” nodes • Need a good measure of purity! – Maximal when? – Minimal when? 14

15. Information Gain Which test is more informative? Split over whether Split over whether Balance exceeds 50K applicant is employed Less or equal 50K Over 50K Unemployed Employed 15

16. Information Gain Impurity/Entropy (informal) – Measures the level of impurity in a group of examples 16

17. Impurity Very impure group Less impure Minimum impurity 17

18. Entropy: a common way to measure impurity • Entropy = i pi log 2 pi pi is the probability of class i Compute it as the proportion of class i in the set. 16/30 are green circles; 14/30 are pink crosses log2(16/30) = -.9; log2(14/30) = -1.1 Entropy = -(16/30)(-.9) –(14/30)(-1.1) = .99 • Entropy comes from information theory. The higher the entropy the more the information content. What does that mean for learning from examples? 18

19. 2-Class Cases: Minimum • What is the entropy of a group in which impurity all examples belong to the same class? – entropy = - 1 log21 = 0 not a good training set for learning • What is the entropy of a group with Maximum 50% in either class? impurity – entropy = -0.5 log20.5 – 0.5 log20.5 =1 good training set for learning 19

20. Information Gain • We want to determine which attribute in a given set of training feature vectors is most useful for discriminating between the classes to be learned. • Information gain tells us how important a given attribute of the feature vectors is. • We will use it to decide the ordering of attributes in the nodes of a decision tree. 20

21. Calculating Information Gain Information Gain = entropy(parent) – [average entropy(children)] child parent entropy entropy Entire population (30 instances) 17 instances child entropy Weighted) Average Entropy of Children = 13 instances 17 13 0.787 0.391 0.615 30 30 nformation Gain= 0.996 - 0.615 = 0.38 for this split 21

22. Entropy-Based Automatic Decision Tree Construction Training Set S Node 1 x1=(f11,f12,…f1m) What feature x2=(f21,f22, f2m) should be used? . What values? . xn=(fn1,f22, f2m) Quinlan suggested information gain in his ID3 system and later the gain ratio, both based on entropy. 22

23. Using Information Gain to Construct a Decision Tree 1 Choose the attribute A Full Training Set S Attribute A with highest information gain for the full training 2 v1 v2 vk set at the root of the tree. Construct child nodes for each value of A. Set S S={ssS | value(A)=v1} Each has an associated subset of vectors in 3 which A has a particular repeat value. recursively till when? 23

24. Simple Example Training Set: 3 features and 2 classes X Y Z C 1 1 1 I 1 1 0 I 0 0 1 II 1 0 0 II How would you distinguish class I from class II? 24

25.X Y Z C 1 1 1 I 1 1 0 I 0 0 1 II 1 0 0 II Eparent= 1 Split on attribute X If X is the best attribute, this node would be further split. II X=1 II Echild1= -(1/3)log2(1/3)-(2/3)log2(2/3) I I = .5284 + .39 II II II = .9184 X=0 Echild2= 0 GAIN = 1 – ( 3/4)(.9184) – (1/4)(0) = .3112 25

26.X Y Z C 1 1 1 I 1 1 0 I 0 0 1 II 1 0 0 II Eparent= 1 Split on attribute Y I I Y=1 Echild1= 0 I I II II Y=0 II II Echild2= 0 GAIN = 1 –(1/2) 0 – (1/2)0 = 1; BEST ONE 26

27.X Y Z C 1 1 1 I 1 1 0 I 0 0 1 II 1 0 0 II Eparent= 1 Split on attribute Z I Z=1 II Echild1= 1 I I II II Z=0 I II Echild2= 1 GAIN = 1 – ( 1/2)(1) – (1/2)(1) = 0 ie. NO GAIN; WORST 27

28. Portion of a fake training set for character recognition Decision tree for this training set. What would be different about a real training set?

29.feature vector class Try the shape feature (square, red) I (square, blue) I (circle, red) II (circle blue) II I I I I (triangle, red) I Entropy? (triangle, green) I II II II II (ellipse, blue) II (ellipse, red) II square ellipse circle triangle I I II II I I II II Entropy? Entropy? Entropy? Entropy? GAIN? 29