Answer Happy

(a) (2 points) In Q-Learning, what are the two possible ways an action given the current state can be selected? (b) (2 points) Give a reason why value iteration is slower than policy iteration? (c) (2 points) Given the alarm Bayesian network, the equation P(B = b|A = a, E = e) < P(B = b|A = a) is correct given an assumption. What is that assumption? (d) (2 points) Draw a Bayesian network for a uni-gram language model, where each word is not affected by any of the n words in the sequence. (e) (2 points) For Laplace smoothing during probability calculation, as we increase the , the probabilities tend to become closer to distribution. (f) (2 points) What is the difference between syntax and semantics in propositional logic? Give an example for each. (g) (2 points) Given two logic formulas a and b, when will the following be False? i) a Vb ii) a + b iii) a b iii) a Vb (h) (2 points) Given a knowledge base KB and a logical formula g, show how entailment relate to contradiction? (i) (2 points) Convert the following formula to Conjunctive Normal Form (CNF). (A + B) → C. ii) A+ ( BC). 6) (2 points) What will be the conclusion if modes ponens is applied to the following premises. Watch(John, Avengers) Act (Avengers, RDJ) Va Vb Vc Watch(a, b) A Act(b, c) + Likes(a,c)