aq algorithm is a different design approach taken by the standard algorithms el few random bits are added to a part of their logic. They are different from determin- ‘istic algorithms; that follow a definite procedure to get the same output every time an input passed whereas randomized algorithms produce a different output every time they're INPUT OUTPUT ALGORITHM Figure 11.5: Deterministic ApproachCOMPUTATIONAL APPROACHES TO PROBLEM SOLVING a epecin inc ee acest a ton i agate et parts. Cons of Randomized Approach « In the worst case, a randomized algorithm may be very slow. : «+ There is a finite probability of getting incorrect answer. However, the probability of getting a wrong answer can be made arbitrarily small by the repeated employment of randomness. * Getting true random numbers is almost impossible. ‘Types of Randomized Algorithms 1. Las Vegas: It is a randomized algorithm that always returns a correct result. But the ‘running time may vary between executions. Example: Randomized QUICK SORT Algorithm 2. Monte Carlo: Itis a randomized algorithm that terminates in polynomial time, but might produce erroneous result. Example: Randomized MINCUT Algorithm Example 1: A company selling jeans gives a coupon with each jeans. There are n different coupons. Collecting n different coupons would give you a free jeans. How many jeans do you expect to buy before you get a free jeans? * The coupon collection process is in phase j when you have already collected j different coupons and are buying to get a new type. * Anew type of coupon ends phase j and you enter phase j + 1. + LetXj be the random variable equal to the number of jeans you buy in phase j. + Then,alculate mt Lx = is the number of jeans bought to har i we n different coupons, ‘The expected number of jeans bought in phase j,randomized Approach . succes i ‘The success probability, p in the j-th phase is E[Xjlis: ‘qhe expected number of jeans bought before all n types of coupons are collected is E[X] = nH, = (nlogn) Example Suppose a company is giving out coupons with every pair of {ifferent coupons. Once you collect all 5 different coupons, We want to calculate the expected number of jeans you need to coupons. Step-by-Step Breakdown jeans sold, and there are n=5 ‘you get a free pair of jeans. ‘buy to collect all 5 different 1, First Coupon: + You buy your first pair of jeans and get your first coupon. This is always new since you don't have any coupons yet. + Expected number of jeans bought so far: 1. 2. Second Coupon: + "The probability of getting a new coupon is 4; coupons are still new to you. + The expected number of jeans t0 get a new coupon is 5/4. {Expected number of jeans bought s0 far: 145/4=2.25. 3. Third Coupon: «The probability of getting a new coupons 3/5 because 3 out oftthe 5 possible coupons are still new to you. «The expected number of jeans to get a MeN CONE is 513. + Expected number of jeans bought so far: 2.25+5/3=2.25+1.67=3.92 4, Fourth Coupon: » ‘The probability of geting anew coupon 1S 4I5 because 2 out of the 5 possible coupons are still new to you: + The expected number of jeans to Bet ¢ «+ Expected number of jeans ‘bought so far: 5, Fifth Coupon: ‘i The probability of getting coupons is still new to you. '5 because 4 out of the 5 possible ew coupon is 5/2. - 3,9245/2=3.9242. 5=6.42. anew coupon is 1/5 because 1 out ofthe 5 possible11, COMPUTATIONAL APPROACHES TO PROBLEM SOLVING 205 __Chopl 11. COMPUTATION ted number of jeans to get a new cou zy Tas eros 6A24+5=11.42, or of jeans bought so far: 6. pees Peete jeans ae need to buy to collect all 5 different coupons 0, the expected BRpromatly 1142 General Formula Application ‘To verify using the formula spon is S/I=5. E(n) ="-Hy Forn=5, the harmonic number HS is: HS=141/2413+1/4+1/5=2.283 Thea E(5)=5*2.283=11.42 Example 2: n people go to a party and drop off their hats to a hat-check person. ‘When the party is over, a different hat-check person is on duty and returns the n hats randomly back to each person. What is the expected number of people who get back their hats? Let X be the random variable representing the number of people who get their own hat back. + Forn=3, find E[X] by first computing the probability mass function Px, and then applying the definition of expectation. * Find a general formula for F(X], for any positive integer n, To find the expected number of people who get their own hat back for n=3,we'll first Compute the probability mass function (PMF) P(X=k) for k=0,1,2,3 and then apply the definition of expectation. Step-by-Step Calculation 1. Enumerate all Possible Permutations: For n=3, there are 3!=6 possible permutations of the hats. We list them: (1,2,3), 13,2), (2,1,3),(2,3,1),3,1,2) and (3,2,1) 2. Count Fixed Points in Each Permutation: + (1,2,3): All 3 people All 3 people get their own hat back G3 fixed points), + G1, 3,2): Only person 1 gets their hat back (1 fixed point), * @, 1,3); Only person 3 gets their hat back (1 fixed point), * 23,1): No one gets their own hat back (0 fixed points). * G,1,2): No one gets their own hat back (0 fixed points), * G,2, 1): Only person 2 gets their hat back (1 fixed point).NsowING =| nt coupons i 2,3), 118 Randomized Approach 207 3 ome Probability Mass Funetion POX=t) + PEK=0): Ni si 5 ): Number of permutations with 0 fixed points / Total permutations = 216 = 1/3 + PCX=1): Number of a f Permutations with 1 fixed point / Total permutations = 3/6 £ POR2): Number of permutations with 2 fixed points ° + P(K=3): Number of, 4, Definition of eee tutions wih 3 fixed pois = 16 E(x] = Yerx ) ca) 5. Compute Expectation for n=3: E[X]-0*P(X=0)+1*P(X=1)+2*P(X=2)+3*P(X=3} =I S =3) E[X]= 0° 1/3+ 11/2424013"1/6 E[X]=0+ 1/2+0+1/2= 1 ‘The expected number of people who get their own hat back for n=3 is E[X]=1. ‘The expected number of people who get their own hat back can be found using indicator random variables and the linearity of expectation. For i = 1,...,7, let X; be the indicator variable that equals 1 if person i gets their hat back, and 0 otherwise. 1 if person i gets their hat back 0 otherwise Let X be the total number of people who get their own hat back: X= x] By the linearity of expectation: EX]=E [e+] LEX mi | mt Each person has a probability of + of getting their own hat back: x E(x] = P(X = Therefore EX]= ‘Tins, the expected number of people who get their own hat back is 1.Es TO PROBLEM SOLVING, CH 208 _Chapter 11 COMPUTATIONAL APPROAS si roach rhe Renda er 11.7. Motivations algorithms offer several significam : oe simpler: For instance, finding the kth using deterministic methods, icked to parti- ized algorithms are often SiMP © Seas .d list can be complex ae hh where a random element is I rithm. Randomize! 1, Simplicity: Rando’ ‘smallest element in an unordere However, using a randomized approacl tion the problem, results in a much simpler alB0 ‘i better asymptotic running times. For iency: ced algorithms can have mpto i pemeraca, sore algorithms run in exponential time, while randomized algorithms run in polynomial time. zation is useful when dealing with incomplete infor. ice and Bob, each with an n-bit number. the same with minimal communica. ging n bits, but a randomized protocol h probability using only O(logn) bits, 3. Lack of Information: Randomi: f ‘mation. For example, consider two parties, Ali re They need to determine if their numbers tion. A deterministic protocol requires exchan} based on fingerprinting can achieve this with hig! 4, Symmetry Breaking: Randomization helps in designing contention resolution mech- anisms in distributed protocols. For example, the Ethernet protocol uses random back-off duration to manage simultaneous transmission attempts effectively, without prior communication. . Counting via Sampling: Randomization can help estimate the size of large spaces or sets through sampling. For instance, computing the integral of a multivariate function over a region can be approximated by sampling points from a bounding box and determining the fraction that lies below the function. 6. Searchit : ii i se aaa pow Randomized algorithms are effective in finding witnesses cee een ties ae the density of witnesses is high. In Polynomial Identity a iple, random points are likel identi Boi aia sno sdncahges ly to identify non-zero evaluations if the