This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: EE 351K PROBABILITY & RANDOM PROCESSES FALL 2011 Instructor: Sujay Sanghavi [email protected] Homework 8 Solution Problem 1 Nefeli, a student in a probability class, takes a multiplechoice test with 10 questions and 3 choices per question. For each question. there are two equally likely possibilities, independent of other questions: either she knows the answer, in which case she answers the question correctly. or else she guesses the answer with probability of success 1/3. (a) Given that Nefeli answered correctly the first question, what is the probability that she knew the answer to that question? (b) Given that Nefeli answered correctly 6 out of the 10 questions, what is the posterior PMF of the number of questions of which she knew the answer? Sol : (a) It is easy to get that P ( know answer  answer is correct ) = P ( answer is correct  know answer ) P ( know answer ) P ( answer is correct ) = 1 2 × 1 1 2 × 1 + 1 3 × 1 2 = 3 4 . (b) Let random variable M to be the number of problems that Nefeli knew the answer. Let E be the event that 6 of answers are correct. Then we have M is binomial distributed with parameter (10 , 1 / 2) , so P ( M = m ) = ( 10 m ) ( 1 2 ) m ( 1 2 ) 10 m for ≤ m ≤ 10 . If given M = m , the conditional probability of E is given by P ( E  M = m ) = { m > 6 , ( 10 m 6 m ) ( 1 3 ) 6 m ( 2 3 ) 4 ≤ m ≤ 6 . So using total probability formula , we will have P ( E ) = 6 ∑ m =0 P ( E  M = m ) P ( M = m ) = ( 10 4 ) ( 2 3 ) 6 ( 1 3 ) 4 . Then using Bayes rule, we will have P ( M = m  E ) = { m > 6 , ( 6 m ) ( 3 4 ) m ( 1 4 ) 6 m ≤ m ≤ 6 . Also, you can get this PMF directly from the result that for every problem, the probability of knowing the answer given the answer is correct is given by 3/4 in part (a), so what we want is just binomial distributed with parameter (6 , 3 / 4) . Problem 2 Suppose points in R are being obtained from two classes, C 1 and C 2 , both of which are normally distributed with parameters (1 , 1) and ( 1 , 1) respectively. If it is known that the priors of C 1 and C 2 are 1/5 and 4/5 respectively, what is the Bayesian optimal decision boundary? Sol : We know that f X  C 1 ( x ) = 1 √ 2 π e ( x 1) 2 2 , f X  C 2 ( x ) = 1 √ 2 π e ( x +1) 2 2 , P ( C 1 ) = 1 / 5 , P ( C 2 ) = 4 / 5 . So for the Bayesian boundary, we have 1 5 √ 2 π e ( x 1) 2 2 = 4 5 √ 2 π e ( x +1) 2 2 ....
View
Full Document
 Spring '07
 BARD
 Conditional Probability, Probability theory, dθ, LMS estimator, Linear LMS Estimator

Click to edit the document details