3 Divide and Conquer The most well known algorithm design strategy: Divide instance of problem into two or more smaller instancesSolve smaller instances recursivelyObtain solution to original (larger) instance by combining these solutions
4 Divide-and-conquer technique a problem of size nsubproblem 1of size n/2subproblem 2of size n/2a solution tosubproblem 1a solution tosubproblem 2a solution tothe original problem
5 Divide and Conquer Examples Sorting: mergesort and quicksortTree traversalsBinary searchMatrix multiplication-Strassen’s algorithmConvex hull-QuickHull algorithm
6 T(n) = aT(n/b) + f (n) where f (n) ∈ Θ(nk) a < bk T(n) ∈ Θ(nk) General Divide and Conquer recurrence: Master TheoremT(n) = aT(n/b) + f (n) where f (n) ∈ Θ(nk)a < bk T(n) ∈ Θ(nk)a = bk T(n) ∈ Θ(nk lg n )a > bk T(n) ∈ Θ(nlog b a)Note: the same results hold with O and Ω.This is the same slide from chapter 2 (analysis of recursive algorithms)It is a good idea to remind students at this point and do some examples especiallyif you went over this quickly and did not do examples the previous time.Here are examples from before:T(n) = T(n/2) + 1T(n) = 2T(n/2) + nT(n) = 3 T(n/2) + n=> this last example is important to mention at some point:Θ(nlog23)NB: cannot “eliminate” or simplify this constant
7 Mergesort Mergesort is a recursive sort algorithm. Algorithm: Split array A[1...n] in two and make copies of each half in arrays B[1... n/2 ] and C[1... n/2 ]Sort arrays B and C
8 Mergesort Merge sorted arrays B and C into array A as follows: Repeat the following until no elements remain in one of the arrays:compare the first elements in the remaining unprocessed portions of the arrayscopy the smaller of the two into A, while incrementing the index indicating the unprocessed portion of that arrayOnce all elements in one of the arrays are processed, copy the remaining unprocessed elements from the other array into A.
9 Mergesort ExampleGo over this example in detail, then do another example of merging, something like:( )(3 4 6)
10 Mergesort … Algorithm Mergesort (A[0 … n – 1]) //sorts array A[0 … n – 1] by recursive mergesort//Input: an array A[0 … n – 1] of orderable elements//Output: array A[0 … n – 1] sorted in nondecreasing orderIf n > 1copy A[0 …n/2] to B[0 …n/2 – 1]copy A[n/2 … n - 1] to C[0 …n/2 – 1]Mergesort (B[0 …n/2 – 1])Mergesort (C[0 …n/2 – 1])Merge(B, C, A)
11 Efficiency of mergesort C(n) = 2C(n/2) + Cmerge(n) for n > 1, C(1) = 0The worst case:W(n) = 2W(n/2) + n – 1 for n > 1, W(1) = 0even if not analyzing in detail, show the recurrence for mergesort in worst case:T(n) = 2 T(n/2) + (n-1)worst case comparisons for merge
12 Efficiency of mergesort All cases have same efficiency: Θ( n log n)Number of comparisons is close to theoretical minimum for comparison-based sorting:log n ! ≈ n lg n nSpace requirement: Θ( n )Can be implemented without recursion (bottom-up)even if not analyzing in detail, show the recurrence for mergesort in worst case:T(n) = 2 T(n/2) + (n-1)worst case comparisons for merge
13 Efficiency of Mergesort Merge analysis (McConnell):Case 1: where all of the elements of list A are smaller than the first element of list B.Case 2: what if the first element of A is greater than the first element of B but all of the elements of A are smaller than the second element of B?Case 3: consider what happens if the elements of A and B are “interleaved” based on their value. In other words, what happens if the value of A is between B and B, the value of A is between B and B, the value of A is between B and B, and so on.Which is the best case? Worst case?
14 ExerciseA. Write a pseudocode for a divide-and-conquer algorithm for the exponentiation problem of computing an where a > 0 and n is a positive integer.B. Set up and solve (for n = 2k) a recurrence relation for the number of multiplications made by this algorithm.C. How does this algorithm compare with the brute-force algorithm for this problem?
15 Quicksort Select a pivot (partitioning/pivot element) Rearrange the list so that all the elements in the positions before the pivot are smaller than or equal to the pivot and those after the pivot are larger than the pivot (See algorithm Partition/PivotList in section 4.2)Exchange the pivot with the last element in the first (i.e., ≤ sublist) – the pivot is now in its final positionSort the two sublistspA[i]≤pA[i]>p
16 Quicksort algorithm Algorithm Quicksort (A[l … r]) If l < r s ← partition (A[l … r]) // s is a split positionQuicksort (A[l … s – 1])Quicksort (A[s + 1 … r])
18 Quicksort …An efficient method based on two scans of the subarray used in the algorithm.Three situations may arise:1) if scanning indices i and j have not crossed, i.e., i < j, we simply exchange A[i] and A[j] and resume the scans.2) if the scanning indices have crossed over, i.e., i > j, we have partitioned the array after exchanging the pivot with A[j]. (n + 1 comparisons)3) if the scanning indices stop while pointing to the same element, i.e., i = j, the value they are pointing to must be equal to p. Thus, the array is partitioned. (n comparisons)
20 Efficiency of Quicksort PivotList analysis:Case 1: when PivotList creates two parts that are the same size.Case 2: when the lists are of drastically different sizes. For example, the largest difference in the size of these two lists occurs if the PivotValue is smaller (or larger) than all of the other values in the list. In that case, we wind up with one part that has no elements and the other that has N – 1 elements. What is the number of comparisons?W(N) = ?Which is the best case? Worst case?
21 Efficiency of quicksort Best case: split in the middle — Θ( n log n) (why?)Worst case: sorted array! — Θ( n2) (why?)Average case: random arrays — Θ( n log n) (why?)Improvements:better pivot selection: median of three partitioning avoids worst case in sorted filesswitch to insertion sort on small subfileselimination of recursionthese combine to 20-25% improvementConsidered the method of choice for internal sorting for large files (n ≥ 10000)
22 Binary Search 1. Inspect the middle item of an array of size N. Binary search is an efficient algorithm for searching in a sorted array.Search requires the following steps:1. Inspect the middle item of an array of size N.2. Inspect the middle of an array of size N/2.3. Inspect the middle item of an array of size N/power(2,2) and so on until N/power(2,k) = 1.This implies k = log2Nk is the number of partitions.
23 Binary Search Requires that the array be sorted. Rather than start at either end, binary searches split the array in half and works only with the half that may contain the value.This action of dividing continues until the desired value is found or the remaining values are either smaller or larger than the search value.
24 Binary Search BinarySearch (List, target, N) //list: the elements to be searched//target: the value being searched for//N: the number of elements in the listStart = 1End = NWhile start <= end doMiddle = (start + end) /2Select (compare (list [middle], target)) fromCase -1: strat = middle + 1Case 0: return middleCase 1: end = middle – 1End selectEnd whileReturn 0
25 Binary Search Suppose that the data set is n = 2k - 1sorted items. Each time examine middle item. If larger, look left, if smaller look right.Second ‘chunk’ to consider is n/2 in size, third is n/4, fourth is n/8, etc.Worst case, examine k chunks. n = 2k – 1 so, k = log2(n + 1). (Decision binary tree: one comparison on each level)Best case, O(1). (Found at n/2).Thus the algorithm is O(log(n)).Extension to non-power of 2 data sets is easy.
26 Binary Search Best case : O(1) Worst case : O(log2N) (why?) Average Case : O(log2N)/2 = O(log2N) (why?)
27 Strassen’s matrix multiplication Strassen observed  that the product of two matrices can be computed as follows:C00 C A00 A B00 B01= *C10 C A10 A B10 B11M1 + M4 - M5 + M M3 + M5=M2 + M M1 + M3 - M2 + M6
29 Efficiency of Strassen’s algorithm If n is not a power of 2, matrices can be padded with zerosNumber of multiplications:Number of additions:Other algorithms have improved this result, but are even more complex
30 Convex-Hull ProblemThere are several divide-and-conquer algorithms for the convex-hull problem of finding the smallest convex polygon that contains n given points in the plane.
31 QuickHull Algorithm Inspired by Quicksort to compute Convex Hull: Assume points are sorted by x-coordinate valuesIdentify extreme points P1 and P2 (part of hull)PmaxP2P1
32 QuickHull Algorithm Compute upper hull: find point Pmax that is farthest away from line P1P2If there is a tie, the point that maximizes the angle P1 PmaxP2compute the hull of the points to the left of line P1Pmaxcompute the hull of the points to the right of line PmaxP2Compute lower hull in a similar manner.PmaxP2P1
33 Geometric OperationsIf p1 = (x1, y1), p2 = (x2, y2), and p3 = (x3, y3) are three arbitrary points in the plane, then the area of the triangle Δp1 p2 p3 is equal to one half of the magnitude of the determinant.|x1 y1 1||x2 y2 1| = x1 y2 + x3 y1 + x2 y3 – x3 y2 – x2 y1 – x1 y3|x3 y3 1|The sign of this expression is positive if and only if the point p3 = (x3, y3) is to the left of the line p1 p2 .The formula can be used to check whether a point lies to the left of a line determined by two other points as well as find the distance from the point to the line.
34 Efficiency of QuickHull algorithm Finding point farthest away from line P1P2 can be done in linear time.This gives same efficiency as quicksort:Worst case: Θ( n2)Average case: Θ( n log n)If points are not initially sorted by x-coordinate value, this can be accomplished in Θ( n log n) — no increase in asymptotic efficiency classOther algorithms for convex hull:Graham’s scan (Θ( n log n) )DCHull (Θ( n log n) )
35 Sequential (Linear) Searching Sequential search – starts at the beginning and examines each element in turn.If we know the array is sorted and we know the search value, we can start the search at the most efficient end.If the array is sorted, we can stop the search when the condition is no longer valid (that is, the elements are either smaller or larger than the search value).
36 Sequential SearchSequential Search: examine each piece of data until the correct one is found.1. Worst case: Order O(n).2. Best case: O(1)3. Average case: O(n/2).Thus, we say sequential search is O(n).