Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Exercise 2 – Probability Math tools November 2, 2015 1 Independence a. Let X1 , . . . , Xn be independent discrete random variables. Show using the definition of independence that: E[X1 X2 ] = E[X1 ]E[X2 ], b. Let (Ω1 , P 1), . . . , (Ωk , Pk ) be distributions, and for every i ∈ [k] let Xi : Ωi → U be a random variable. Let (Ω, P ) = (Ω1 , P 1) × · · · × (ω, Pk ). For every i ∈ [k] we define Xi0 : Ω → U to by ∀ω = (ω1 , . . . , ωk ) ∈ Ω, Xi0 (ω) = Xi (ωi ) Show that {Xi0 }ki=1 are independent. c. Let {Ai }ni=1 be a set of independent events. Show that if we replace a subset I ⊆ [n] with their complements the set is still independent. Namely, that for every I ⊆ [n] the set {Ai }i∈I / ∪ Āi i∈I is independent. 2 Bayes’ theorem Let A, B ∈ Ω be two events. Assuming P r(B) 6= 0, define the probability of A conditioned on B as P r(A ∩ B) P r(A|B) = . P r(B) The intuitive interpretation of this number is the probability that A occurs given that B has occurred. a. Proof Bayes’ theorem, namely that P r(A|B) = P r(A)P r(B|A) , P r(A)P r(B|A) + P r(A)P r(B|A) where A is the complement of A. 1 b. A card is drawn from a standard deck of 52 cards and discarded (i.e. not replaced). A second card is drawn from the remaining deck of 51 cards. Given that the second card was a spade, what is the probability that the first card was also a spade? c. A medical company touts its new test for a certain genetic disorder. The false negative rate is small: if you have the disorder, the probability that the test returns a positive result is 0.98. The false positive rate is also small: if you do not have the disorder, the probability that the test returns a positive result is only 0.04. Assume that 2% of the population has the disorder. If a person chosen uniformly at random from the population is tested and the result comes back positive, what is the probability that the person has the disorder? 3 Phase transitions in random graphs In this question we consider a few examples of a phase transition in a random graphs. Roughly speaking, phase transition is a phenomenon of abrupt change in the behavior of an object when some parameter passes a certain threshold. We look at a random graph G ∼ G(n, p). Definition 3.1. Let p : N → (0, 1) be a function. We say that a property P of a graph has a threshold at p if for any function p0 : N → (0, 1) it holds that n→∞ • If p0 = cp with c > 1 then PrG∼G(n,p0 (n)) [G has the proerty P ] −→ 1. n→∞ • If p0 = cp with c < 1 then PrG∼G(n,p0 (n)) [G has the proerty P ] −→ 0 Show that the property of having no isolated vertexes has a threshold at p= ln n n Namely: n→∞ a. if p0 = c lnnn , c > 1 then Pr [G has no isolated vertexes] −→ 1 n→∞ b. if p0 = c lnnn , c < 1 then Pr [G has no isolated vertexes] −→ 0 4 Stock Pricing The stock price of googface behaves the following way: on each day its price either gains 60% or looses 40% compared to its price of the previous day, and the two events happen with probability 1/2 each, independently from historical price movements. Let the random variable Xi denote the price of the stock on i-th day, where X0 = 1. a. Compute E[Xi ] for all i. b. Compute E[ln(Xi )] for all i. 2 c. Compute var[ln(Xi )] for all i (hint: write ln Xi as a sum of independent variables). d. Suppose you buy the stock at day 0 and sell it on day 10, 000. Use Chebyshev’s inequality to bound the probability that you do not lose. 5 The Binomial distribution Recall the binomial distribution from the tirgul - We flip n independent coins, where each comes out heads with probability p. Let X be the total number of heads. The distribution of X is called a binomial distribution with parameters n and p, and denoted X ∼ B(n, p). ) ≤ n4 . Here we prove a for a similar case We saw in the tirgul that for p = 12 - Pr(X ≥ 3n 4 that the bound is actually exponentially small in n. a. Use Chebyshev’s inequality to show that if p = c for some constant c ∈ (0, 1/2), then the probability that X > n2 goes to zero as n goes to infinity. b. Prove by an elementary direct computation (without using Stirling’s formula or any other heavy machinery) that the above probability is exponentially small in n. (Hint: start by bounding the probability that X = n2 , and show that any higher 1 1 value isn even less probable. Also, remember that ∀p ∈ (0, 2 ) p(1 − p) < 4 and n < 2 ). 2 3