This preview shows pages 1–4. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 6.042/18.062J Mathematics for Computer Science Tom Leighton and Ronitt Rubinfeld December 20, 2006 Final Final Problem 1. [8 points] Prove that for all n N, the following identity holds
n i=1 2 i2 = n(n + 1)(2n + 1) . 6 Solution. By induction on n 1, with induction hypothesis P (n) : for all n N Base case (n = 1): 1(1 + 1)(2 + 1) 6 = = 1 = 12 6 6 Inductive step: Assume P (n), we need to show that P (n + 1) holds.
n i=0 i2 = n(n + 1)(2n + 1) 6 n+1 i=0 n i2 = ( i2 ) + (n + 1)2 i=0 n(n + 1)(2n + 1) = + (n + 1)2 6 n(n + 1)(2n + 1) + 6(n + 1)2 = 6 (n + 1)(2n2 + n + 6n + 6) = 6 2 (n + 1)(2n + 7n + 6) = 6 (n + 1)(n + 2)(2n + 3) = 6 (n + 1)((n + 1) + 1)(2(n + 1) + 1) = 6 P (n + 1) as required. We have shown that P (n) P (n + 1). Thus, P (n) is true for all n N. Final 3 Problem 2. [20 points] CoinFlip is a 2 player game. Each player wins with probability exactly 0.5. There are no ties. n people are playing a CoinFlip tournament. Every person plays a CoinFlip game with every other person exactly once. Thus everybody plays n  1 games. The outcomes of all the games are mutually independent of one another. We say that the tournament is a success if for every i {0, 1, . . . , n  1}, there is exactly one player, which we will refer to as pi , with exactly i wins. (a) [10 points] Prove that if the tournament is a success, then for any integers j, k with 0 k < j n  1, pj defeats pk . Solution. We prove it by induction on k. The inductive hypothesis P (k) is that for all 0 k and all < j < n, pj defeats p . The base case is k = 0. Now p0 loses all n  1 games. Thus, for every j > 0, pj defeats p0 . Suppose P (k) and let us show P (k + 1). Since the tournament is a sucess, pk+1 wins exactly k + 1 games. Because P (k) holds, pk+1 defeats the k + 1 players p0 , p1 , . . . , pk . Thus, for all k + 1 < j < n, pj defeats pk+1 . So P (k + 1) holds. (b) [6 points] What is the probability that the tournament will be a success? Solution. Let be a permutation of {0, 2, . . . , n  1} and define the event E to be that the tournament is a success with players pi winning exactly i games, where pi is the (i)th player. Then the events E and E are disjoint for = . Moreover, by symmetry, Pr[E ] = Pr[E ] for all permutations , . Let E be the event the tourna ment is a success. Since the events are disjoint, Pr[E] = Pr[E ] = n! Pr[E ], where is some fixed permutation. Now E determines all the outcomes of the games. As there are n games, we have 2 (n) 1 2 Pr[E ] = . 2 In total, (n) 1 2 Pr[E] = n! . 2 (c) [4 points] Show that your answer to part (b) is o(1). Solution. We have, (n) (n) (n)n log n 1 2 1 2 1 2 n log n Pr[E] n =2 = . 2 2 2
n Final 4 Now, n  n log n = (n2 ), so there is a positive constant c such that for sufficiently 2 large n, this probability is at most cn2 1 , 2 which for sufficiently large n, is clearly less than any positive constant, and thus is o(1). Final 5 Problem 3. [8 points] A person is passing time by advancing a token on the set of natural numbers. In the beginning, a token is placed on 0. The person keeps playing moves forever. Each move proceeds as follows: 1. First the person tosses a fair coin (with heads/tails equally likely). 2. Suppose the token is currently placed on n. If heads came up, then the person moves the token to n + 3, otherwise he moves the token to n + 4. For each n N, let En be the event "There was a move on which the token landed on n". Let pn = Pr[En ]. Find a recurrence relation for pn . You do not need to solve the recurrence, but you should specify the boundary conditions that would be necessary to find a solution to the recurrence. 1 1 pn = pn3 + pn4 , 2 2 with boundary conditions p0 = 1, p1 = 0, p2 = 0, p3 = 1/2. Solution. For all n 4, Final Problem 4. [10 points] Exactly 1/5th of the people in a town have Beaver Fever . 6 There are two tests for Beaver Fever, TEST1 and TEST2. When a person goes to a doctor to test for Beaver Fever, with probability 2/3 the doctor conducts TEST1 on him and with probability 1/3 the doctor conducts TEST2 on him. When TEST1 is done on a person, the outcome is as follows: If the person has the disease, the result is positive with probability 3/4. If the person does not have the disease, the result is positive with probability 1/4. When TEST2 is done on a person, the outcome is as follows: If the person has the disease, the result is positive with probability 1. If the person does not have the disease, the result is positive with probability 1/2. A person is picked uniformly at random from the town and is sent to a doctor to test for Beaver Fever. The result comes out positive. What is the probability that the person has the disease? Solution. Let B be the event that the person has BLAH. Let T 1 be the event that the person is tested with test1. Let T 2 be the event that the person is tested with test2. Let P be the event that the test comes out positive. A tree diagram is worked out below with the given information: Final 7 + BLAH 1/5 _ 3/4 1/4 1/10 1/30 2/3 Test1 No BLAH 4/5 + 3/4 2/15 _ 1/4 2/5 1 1/15 + _ + 0 0 Test2 BLAH 1/3 No BLAH 1/5 1/2 2/15 1/2 4/5 _ 2/15 Image by MIT OpenCourseWare. Final The probability that a person has BLAH, given that the test comes out positive is: 8 Pr {B  S} = = = Pr {B  T 1 P } Pr (T 1) + Pr {B  T 2 P } Pr (T 2) Pr (B T 1 P ) Pr (B T 2 P ) Pr (T 1) + Pr (T 2) Pr (T 1 P ) Pr (T 2 P ) Pr (D T 1 P ) Pr (T 1) + Pr (D T 1 P ) + Pr D T 1 P Pr (D T 2 P ) Pr (T 2) Pr (D T 2 P ) + Pr D T 2 P
1 10 1 10 = = + 2 15 2 + 3 1 15 1 15 + 2 15 1 3 5 13 Final 9 Problem 5. [10 points] Two identical complete decks of cards, each with 52 cards, have been mixed together. A hand of 5 cards is picked uniformly at random from amongst all subsets of exactly 5 cards. (a) [5 points] What is the probability that the hand has no identical cards (i.e., cards with the same suit and value. For example, the hand Q, 5, 6, 8, Q has iden tical cards.)? We can calculate this probability by computing hands with no identical cards total possible hands There are 104 cards. There are 5 cards in a hand. Order does not matter. The total number of possible hands is: 104 5 There are 52 possible card faces, and we can choose 5 of them if no identical cards are allowed. Additionally, each card can be from either deck 1 or deck 2. Therefore the number of hands with no identical cards, chosen from 2 decks is: 52 25 5 Therefore the probability of drawing a hand with no identical cards is: 52 25 5 104
5 (b) [5 points] What is the probability that the hand has exactly one pair of identical cards? This can be solved by a similar approach. A hand of this type can be distin guished by the face (suit and value) of the repeated card, and by the faces of the 3 nonrepeated cards. There are 52 possible values for the face of the repeated card. There are 51 possible faces for the nonrepeated cards, since none of these can be 3 repeated. Each of these could come from either the 1st deck or the 2nd deck. There are 104 possible hands, as before. So the probability of getting a hand with exactly 5 one pair of identical cards is: 52 51 23 3 104
5 Final 10 Problem 6. [28 points] Scores for a final exam are given by picking an integer uniformly at random from the set {50, 51, . . . , 97, 98}. The scores of all 128 students in the class are assigned in this manner. For parts (a), (b), (c) and (d) you may NOT assume that these scores are assigned independently. For parts (e), (f), (g) and (h) you MAY assume that these scores are assigned independently. 128 1 Let S1 , . . . , S128 be their scores. Let S = 128 ( i=1 Si ) be the average score of the class. (a) [3 points] For i {1, . . . , 128}, what is E[Si ] ? (b) [2 points] Show that E[S] = 74. Make no independence assumptions. (c) [4 points] Prove that Pr[S 88] Make no independence assumptions. 37 . 44 Final 11 (d) [5 points] Improve your previous bound by using the fact that the minimum possible score is 50. Prove that Pr[S 88] Make no independence assumptions. 12 . 19 (e) [4 points] For the remaining problems, assume that all the scores are assigned mutually independently. Use Problem 1 of this final to find V ar[Si ]. Final (f) [3 points] What is V ar[S]? 12 (g) [2 points] What is the standard deviation of S? (h) [5 points] Prove, using the Chebyshev Inequality, that Pr[S 69] 1 . 16 Final Solution. (a) We simply take the average of the numbers from 50 to 98. Thus, E[Si ] = 74. (b) By linearity of expectation,
128 1 1 E[S] = E[ ( Si )] = (128 E[S1 ]) = E[S1 ] = 74 128 i=1 128 50+98 2 13 = (c) By Markov's inequality, Pr[S 88] E[S] 74 37 = = . 88 88 44 (d) We define a random variable T = S  50, and thus E[T ] = E[S]  50 = 24. Now we just apply Markov's inequality: Pr[S 88] = Pr[T 38] (e) We define Ti = Si  50.
48 E[T ] 24 12 = = . 38 38 19 V ar[Si ] = V ar[Ti ] = (f) E[Ti2 ]E2 [Ti ] 1 2 1 (48)(49)(97) =( i )E2 [Ti ] = (24)2 = 776576 = 200. 49 i=0 49 6 128 1 1 2 V ar[S1 ] 200 25 V ar[S] = V ar[ ( Si )] = ( ) (128 V ar[S1 ]) = = = . 128 i=1 128 128 128 16 (g) The standard deviation of S is simply the square root of the variance of S: 25 5 S = = . 16 4 (h) Using Chebyshev's inequality, Pr[S 69] Pr[S  74 5] = Pr[S  E[S] 4 S ] 1 1 = . 16 42 Final 14 Problem 7. [16 points] 1000 files F1 , F2 , . . . , F1000 have just reached a disk manager for writing onto disk. Each file's size is between 0M B and 1M B. The sum of all files' sizes is 400M B. The disk manager has 4 disks under its control. For each file Fi , the disk manager chooses a disk uniformly at random from amongst the 4 disks, and Fi is written to that disk. The choices of disk for the different files are mutually independent. (a) [2 points] What is the expected number of files that will be written to the first disk? We can use indicator variables. For each file, Pi = 1 if Fi is written to the first disk. The chance of an individual file being written to the first disk is 1/4. By linearity of expectation, the expected number of files written to the first disk is the sum of the expected values of Pi 's. The expected value of each indicator variable is 1/4, and ( i=1 1000)1/4 = 250, so the expected number of files to be written to the first disk is 250. (b) [2 points] What is the expected number of bytes written on the first disk? We can say that each file Fi has bit size Si . Each file has a 1/4 chance of being written do the first disk. Therefore, by linearity of expectation, the expected number of bytes written to the first disk is the sum of the expected number of bytes per file written to the first disk, which is:
1000 i=1 1/4 Si = 1/4 1000 i=1 Si = 1/4 400 = 100 Final 15 (c) [8 points] Find the best upper bound you can on the probability that 200M B or more are written on the first disk? For this we can use the first Chernoff bound, which is: Pr (X c Ex (X)) e(c ln c  c + 1) Ex (X) The Chernoff bound only works if X is the sum of random variables that each take on a value between 0 and 1. The file size of each file in the first disk is between 0 and 1Mb . So we can define X to be the total number of bytes in disk 1. The expected value of X is 100, so we take c to be 2. We get: Pr (X 2 100) e(2 ln 2  2 + 1)100 (d) [4 points] Find the best upper bound you can on the probability that there is some disk with 200M B or more written on it? For this we can use the Union Bound along with our result from above. The proba bility of this event happening in one or more disks is upper bounded by the sum of the probabilities of the event happening in each disk. This gives us an upper bound of 4 e(2 ln 2  1)100 MIT OpenCourseWare http://ocw.mit.edu 6.042J / 18.062J Mathematics for Computer Science
Fall 2010 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms. ...
View
Full
Document
This note was uploaded on 01/19/2012 for the course CS 6.042J / 1 taught by Professor Tomleighton,dr.martenvandijk during the Fall '10 term at MIT.
 Fall '10
 TomLeighton,Dr.MartenvanDijk
 Computer Science

Click to edit the document details