Ti = indicator random variable of the event that i-th throw results in a tail E[T] = E[T1] + … + E[T6] = 6*(1/2) = 3 P(T=3) = P(H=3) = binomial(6,3)/26 = 5/16 < 1/2 E[Xi,j] = ½ (consider only i<j) X= Xi,j 1 i<j n E[X] = n(n-1) /4 T = 1 + (1/2) * 0 + (1/2) * ( T + T ) T=1+T There exists c such that T(n) T(n/2)+T(n/3)+c*n. We need to show that there exists d such that T(n) d*n for all n. Induction step: T(n) T(n/2) + T(n/3) + c*n d*n/2 + d*n/3 + c*n d*n + (c-d/6)*n d*n, taking d=6c. l m+1 if B A[m] then Reverse(a,b) for i from a to a+b do swap(A[i],A[a+b-i]); Rotate(k) Reverse(1,k) Reverse(k+1,n) Reverse(1,n) 1,….,k,k+1,….,n k,….,1,k+1,….,n k,….,1,n,….,k+1 k+1,….,n,1,….,k 1) find the median m of A 2) m m m sum S 3) if S\geq C then recurse on A[n/2..n] else recurse on A[1..n/2] with C’=C-S T(n) = T(n/2) + O(n) 3) if S\geq C then recurse on A[n/2..n] else recurse on A[1..n/2] with C’=C-S Coupon collector problem n coupons to collect What is the expected number of cereal boxes that you need to buy? Coupon collector problem Assume that a dart throw is uniform in the circle. Let p be The fraction occupied by the bull’s eye. Expected number of darts needed to hit the bull’s eye ? Coupon collector problem Assume that a dart throw is uniform in the circle. Let p be The fraction occupied by the bull’s eye. Expected number of darts needed to hit the bull’s eye ? 1/p What is the expected number of boxes that I buy in k-th phase ? k-th phase = when I have k different Kinds of coupons. E[X0] = 1 … E[Xk] = ? … E[Xn-1] = n What is the expected number of boxes that I buy in k-th phase ? k-th phase = when I have k different Kinds of coupons. E[X0] = 1 … E[Xk] = n/(n-k) … E[Xn-1] = n What is the expected number of boxes that I buy in k-th phase ? k-th phase = when I have k different Kinds of coupons. n-1 X=X0+X1+…+Xn-1 n n 1 = =n n-k k k=0 k=1 = (n ln n) What is the expected number of boxes that I buy in k-th phase ? k-th phase = when I have k different Kinds of coupons. n-1 X=X0+X1+…+Xn-1 n n 1 = =n n-k k k=0 k=1 E[X]=E[X0]+…+E[Xn-1] = (n ln n) Harmonic numers n ln n 1 k k=1 1+ln n Randomized algorithm for “median” SELECT k-th element for random x 1) <x L =x >x R 2) recurse on the appropriate part Randomized algorithm for “median” Las Vegas algorithm (never makes error, randomness only influences running time) The identity testing algorithm was Monte Carlo algorithm with 1 sided error. Markov inequality For non-negative random variable X: P(X > a.E[X]) < 1/a P(X a.E[X]) 1/a Variance For a random variable X: V[ X ] = E[ (X-E[X])2 ] What is the variance of X=the number on a (6-sided) dice ? Variance For a random variable X: V[ X ] = E[ (X-E[X])2 ] Y = (X-E[X])2 P( Y > a.E[Y] ) < 1/a P( (X-E[X])2 > a.V[X] ) < 1/a P( (X-E[X])2 > b2.E[X]2 ) < V[X]/(b2 E[X]2) P( |X-E[X]| > b.E[X] ) < V[X] E[X]2 * 1 b2 Chebychev’s inequality P( |X-E[X]| > b.E[X] ) < V[X] E[X]2 * P( (1-b)*E[X] X (1+b)*E[X] ) >1 - V[X] E[X]2 * 1 b2 1 b2