VIEWS: 4 PAGES: 57 POSTED ON: 7/25/2011 Public Domain
CS 5100 Artificial Intelligence Prof. C. Hafner Class Notes March 29, 2011 Learning Paradigms • Supervised Learning • Reinforcement Learning (genetic algorithms) • Unsupervised Learning (clustering algorithms) We will consider supervised learning, the most widely used and the most successful paradigm (also the most expensive. Why ??) Supervised Learning Methods • Naïve Bayes • Decision Tree Learning (ID3 and C4.5) • Neural Nets • Support Vector Machines • Naïve Bayes learning – Maximum likelihood estimation – Smoothing • Motivation • How to (simplest method) Information Theory • Information is about categories and classification • We measure quantity of information by the resources needed to represent/store/transmit the information • Messages are sequences of 0’s and 1’s (dots/dashes) which we call “bits” (for binary digits) • You need to send a message containing the identity of a spy – It is known to be Mr. Brown or Mr. Smith • You can send the message with 1 bit, therefore the event “the spy is Smith” has 1 bit of information Calculating quantity of information • Def: A uniform distribution of a set of possible outcomes (X1 . . . Xn) means the outcomes are equally probable; that is, they each have probability 1/n. • Suppose there are 8 people who can be the spy. Then the message requires 3 bits. If there are 64 possible spies the message requires 6 bits, etc. (assuming a uniform distribution) • Def: The information quantity of a message where the (uniform) probability of each value is p: I = -log p bits Intuition and Examples • Intuitively, the more “surprising” a message is, the more information it contains. If there are 64 equally- probable spies we are more surprised by the identity of the spy than if there are only two equally probable spies. • There are 26 letters in the alphabet. Assuming they are equally probable, how much information is in each letter: I = -log (1/26) = log 26 = 4.7 bits • Assuming the digits from 0 to 9 are equally probable. Will the information in each digit be more or less than the information in each letter? Sequences of messages • Things get interesting when we looks beyond a single message to a long sequence of messages. • Consider a 4-sided die, with symbols A, B, C, D: – Let 00 = A, 01=B, 10=C, 11=D – Each message is 2 bits. If you throw the die 800 times, you get a message 1600 bits long That’s the best you can do if A,B,C,D equally probable Non-uniform distributions (cont.) • Consider a 4-sided die, with symbols A, B, C, D: – But assume P(A) = 7/8 and P(B)=P(C)=P(D) = 3/24 – We can take advantage of that with a different code: 0 = A, 10= B, 110 = C, 111 = D – If we throw the die 800 times, what is the expected length of the message? What is the entropy? • ENTROPY is the average information (in bits) of events in a long repeated sequence Entropy Formula for entropy with outcomes x1 . . . xn : - Σ P(xi) * log P(xi) bits For a uniform distribution this is the same as –log P(x1) since all the P(xi) are the same. What does it mean? Consider 6-sided die, outcomes equally probable: -log 1/6 = 2.58 tells us a long sequence of die throws can be transmitted using 2.58 bits per throw on the average and this is the theoretical best Review/Explain Entropy • Let the possible outcomes be x1 . . . . xn – With probabilities p1 . . . pn that add up to 1 • Ex: an unfair coin where n = 2, x1 = H (3/4), x2 = T (1/4) • In a long sequence of events E = e1 . . . ek, we assume that outcome xi will occur k * pi times, etc. E = HHTHTHHHTTHHHHHTHHHHTTHTHTHHHH ……. If k = 10000, we can assume H occurs 7500 times, T 2500. Note: the concept TYPES vs. TOKENS. There are two types and 10000 tokens in this scenario. Review/Explain Entropy The entropy of E H(E) is the average information of the events in the sequence e1 . . . ek : k 1/k * Σ I(ej) = [now switch to summation over outcomes] j=1 n n 1/k * Σ I(xi) * (k*pi) = k/k * Σ I(xi) * pi i=1 i=1 n Σ -log(pi) * pi bits i=1 Review/Explain Entropy Entropy is sometimes called “disorder” – it represents the lack of predictability as to the outcome for any element of a sequence (or set) If a set has just one outcome, entropy = 1 * -log(1) = 0 If there are 2 outcomes, then 50/50 probability gives the maximum entropy – complete unpredictability. This generalizes to any uniform distribution for n outcomes. - (0.5 * log(.5) + 0.5 * log(.5)) = 1 bit Note: log(1/2) = -log(2) = -1 Calculating Entropy • Consider a biased coin: P(heads) = ¾; P(tails) = ¼ • What is the entropy of a coin toss outcome? • H = ¼ * -log(1/4) + ¾ * -log(3/4) = 0.811 bits • Using the Information Theory Log Table • H = 0.25 * 2.0 + 0.75 * 0.415 = 0.5 + 0.311 = .811 • A fair coin toss has more “information” • The more unbalanced the probabilities, the more predictable the outcome, the less you learn from each message. Maximum disorder 1 H (entropy in bits) 0 ½ 1 Probability of x1 Entropy for a set containing 2 possible outcomes (x1, x2) What if there are 3 possible outcomes? for equal probability case: H = -log(1/3) = about 1.58 Define classification tree and ID3 algorithm • Def: Given a table with one result attribute and several designated predictor attributes, a classification tree for that table is a tree such that: – Each leaf node is labeled with a value of the result attribute – Each non-leaf node is labeled with the name of a predictor attribute – Each link is labeled with one value of the parent’s predictor • Def: the ID3 algorithm takes a table as input and “learns” a classification tree that efficiently maps predictor value sets into their results from the table. A trivial example of a classification tree Record# Color Shape Fruit 1 red round apple 2 yellow round lemon 3 yellow oblong banana Color red yellow apple Shape round oblong lemon banana The goal is to create an “efficient” classification tree which always gives the same answer as the table A well-known “toy” example: sunburn data Name Hair Height Weight Lotion Sunburned Sarah Blonde Average Light No Yes Dana Blonde Tall Average Yes No AleX Brown Short Average Yes No Annie Blonde Short Average No Yes Emily Red Average Heavy No Yes Pete Brown Tall Heavy No No John Brown Average Heavy No No Katie Blonde Short Light Yes No Predictor attributes: hair, height, weight, lotion Hair Blonde Brown Red Not Lotion Sunburned Sunburned N Y Not Sunburned Sunburned Outline of the algorithm 1. Create the root, and make its COLLECTION the entire table 2. Select any non-singular leaf node N to SPLIT 1. Choose the best attribute A for splitting N (use info theory) 2. For each value of A (a1, a2, . .) create a child of N, Nai 3. Label the links from N to its children: “A = ai” 4. SPLIT the collection of N among its children according to their values of A 3. When no more non-singular leaf nodes exist, the tree is finished 4. Def: a singular node is one whose COLLECTION includes just one value for the result attribute (therefore its entropy = 0) Choosing the best attribute to SPLIT: the one that is MOST INFORMATIVE that reduces the entropy (DISORDER) the most Assume there are k attributes we can choose. For each one, we compute how much less entropy exists in the resulting children than we had in the parents: H(N) – weighted sum of H(children of N) Each child’s entropy is weighted by the “probability” of that child (estimated by the proportion of the parent’s collection that would be transferred to the child in the split) C(S1) = {S,D,X,A,E,P,J,K}(3,5)/____} S1: _______ Calculate entropy: - [3/8 log 3/8 + 5/8 log 5/8] = .53 + .424 = .954 Find information gain (IG) for all 4 predictors: hair, height, weight, lotion Start with lotion: values (yes, no) Child 1: (yes) = {D,X,K}(0, 3)/0 Child 2: (no) = {S,A,E,P,J}(3,2)/ -[3/5 log 3/5 + 2/5 log 2/5] = .971 Child set entropy = 3/8 * 0 + 5/8 * .971 = 0.607 IG(Lotion) = .954 - .607 = .347 Then try hair color: values (blond, brown, red) Child 1(blond) = {S,D,A,K}(2,2)/1 Child 2(brown) = {X,P,J}(0,3)/0 Child 3(red) = {E}(1,0)/0 Child set entropy = 4/8 * 1 + 3/8 * 0 + 1/8 * 0 = 0.5 IG(Hair color) = .954 - 0.5 = .454 Next try Height: values (average, tall short) Child1(average) = {S,E,J}(2,1)/ -[2/3 log 2/3 + 1/3 log 1/3] = 0.92 Child2(tall) = {D,P}(0,2)/0 Child3(short)={X,A,K}(1,2)/0.92 Child set entropy = 3/8 * 0.92 + 2/8 * 0 + 3/8 * 0.92 = 0.69 IG(Height) = .954 - .69 = 0.26 Next try Weight . . . IG(Weight) 0.954 – 0.94 = 0.014 So Hair color wins: Draw the first split and assign the collections N1: Hair Color Red Blond: C = {S,D,A,K}(2,2)/1 Brown yes no S2:_______ S2:_________ C(S2) = {S,D,A,K}(2,2)/1 Start with lotion: values (yes, no) Child 1: (yes) = {D, K}(0, 2)/0 Child 2: (no) = {S,A}(2,0)/ 0 Child set entropy = 0 IG(Lotion) = 1 – 0 = 1 No reason to go any farther S1: Hair Color Red Blond: C = {S,D,A,K}(2,2)/1 Brown yes no S2: Lotion no yes yes no Discuss assignment 5 Perceptrons and Neural Networks: Another Supervised Learning Approach Perceptron Learning (Supervised) • Assign random weights (or set all to 0) • Cycle through input data until change < target • Let α be the “learning coefficient” • For each input: – If perceptron gives correct answer, do nothing – If perceptron says yes when answer should be no, decrease the weights on all units that “fired” by α – If perceptron says no when answer should be yes, increase the weights on all units that “fired” by α