Basic Probability for Randomized Algorithms


§ Contents

§ Overview

Joint Distribution

Pr⁡[X=x,Y=y] \Pr[X = x, Y = y]

The probability that random variables XX and YY have values xx and yy, respectively.

Conditional Distribution

Pr⁡[X=x ∣ Y=y]=Pr⁡[X=x,Y=y]Pr⁡[Y=y] \Pr[X = x\,|\,Y = y] = \frac{\Pr[X = x, Y = y]}{\Pr[Y = y]}

The probability that XX has value xx when YY has value yy

Marginal Distribution

Pr⁡[X=x]=∑y∈ΩYPr⁡[X=x ∣ Y=y] \Pr[X = x] = \sum_{y \in \Omega_Y} \Pr[X = x\,|\,Y = y]

Independence

Random variables XX and YY are independent if for all x∈ΩXx \in \Omega_X and y∈ΩYy \in \Omega_Y

Pr⁡[X=x]=Pr⁡[X=x ∣ Y=y]\Pr[X = x] = \Pr[X = x\,|\,Y = y]

Boole's Inequality (Union Bound)

Let S1,S2,…,SkS_1, S_2, \dots, S_k be a set of events occuring with probability p1,p2,…,pk.p_1, p_2, \dots, p_k.

The probability that at least one of the events S1,S2,…,SkS_1, S_2, \dots, S_k occurs is at most p1+p2+⋯+pk.p_1 + p_2 + \dots + p_k.

The probability that none of the events S1,S2,…,SkS_1, S_2, \dots, S_k occur is at least 1−(p1+p2+⋯+pk).1 - (p_1 + p_2 + \dots + p_k).

Then,

Pr⁡[A∪B]=Pr⁡[A]+Pr⁡[B]−Pr⁡[A∩B]. \Pr[A \cup B] = \Pr[A] + \Pr[B] - \Pr[A \cap B].