88 Chaprer 4 Divide-and-Conguer 439 Solve the recurrence T(n) = 31) + logn by making a change of variables, Your solution should be asymptotically tight. Do not worry about whether values ave integral method for Although you can use the substitution method to provide a succinct proof that a solution to a recurrence is correct, you might have trouble coming up with a good guess. Drawing out a recursion tree, as we did in our analysis of the merge sort recurrence in Section 2.3.2, serves as a straightforward way to devise a good guess. In a recursion tree, each node represents the cost of a single subproblem somewhere in the set of recursive function invocations. We sum the costs within each level of the tree to obtain a set of per-level costs, and then we sum alll the ‘per-level costs to determine the total cost of all levels of the recursion. A recursion tree is best used to generate a good guess, which you can then verify by the substitution method. When using a recursion tree to generate a good guess, you can ofien tolerate a small amount of “sloppiness,” since you will be verifying your guess later on. If you are very careful when drawing out a recursion tree and summing the costs, however, you can use a recursion tree as a direct proof of a solution to a recurrence. In this section, we will use recursion trees to generate good guesses, and in Section 4.6, we will use recursion trees directly to prove the theorem that forms the basis of the master method. For example, let us see how a recursion tree would provide a good guess for the recurrence Tn) = 37(|n/4|) | @(a2). We start by focusing on finding an upper bound for the solution. Because we know that floors and ceilings usually do not matter when solving recurrences (here’s an example of sloppiness that we can tolerate), We create a recursion tree for the recurrence T(n) = 31 (n/4) | en, having written out the implied constant coefficient ¢ > 0. Figure 4.5 shows how we derive the recursion tree for T(n) = 37(n/4) ~ en. For convenience, we assume that n is an exact power of 4 (another example of tolerable sloppiness) so that all subproblem sizes are integers. Part (a) of the figure shows T(mn), which we expand in part (b) into an equivalent tree representing the recurrence. The cn? term at the root represents the cost at the top level of recursion, and the three subtrees of the root represent the costs incurred by the subproblems of size n/4, Part (c) shows this process carried one step further by expanding each node with cost T(n/4) from part (b). The cost for each of the three children of the root is c(n/4)*. We continue expanding each node in the tree by breaking it into its constituent parts as determined by the recurrence.44 The recursion-iree method for solving recurrences 9 Tin) en /\\ Tie) THE) TR) cay cay J... LS T(z) Te) Tae) Th ©) Tie Tie) @ © © cay en? Ta) TA) TU) TE) TO) TA) TA) TAY TU) TU) TAY TA) TA) O(n) @ Toul: O10") Figure 4.5 Constructing a recursion tree for the recurrence T(n) = 3T(n/4) = en®. Part (a) shows 7 (), which progressively expands in (b}~14) to form the recursion tee. The fully expanded tree in part () has height logy m (it has logy n ~ 1 levels.90 Chaprer 4 Divide-and-Conguer Because subproblem sizes decrease by a factor of 4 each time we go down one level, we eventually must reach a boundary condition. How far from the root do we reach one? The subproblem size for a node at depth i is m/4'. Thus, the ‘subproblem size hits m = I when n/4! = 1 or, equivalently, when i = loggn. Thus, the tree has log, 1" + I levels (at depths 0, 1,2,... .log,m). Next we determine the cost at each level of the tree. Each level has three times, ‘more nodes thun the evel above, and so dhe numer of nodes at depth 7 is 3 se subproblem sizes reduce by a factor of 4 for each level we go down from the root, each node at depth i, for? = 0,1,2,..2,logym — I, has a cost ‘of e(n/4')®, Multiplying, we see that the fotal cost over all nodes at depth i, for i = 0,1.2,...,logen — 1, is 8e(n/4')? = (B/16)'en. The bottom level, at depth log, n, has 3°" — n> nodes, each contributing cost T(1), for a'total cost of n=" 7 (1), which is @(n"*"), since we assume that T(1) is a constant. Now we add up the costs over all levels to determine the cost for the entire tree: T 2 3 on (ZY ent 3 eyghee® 0) = or a eo (B) em anne (Bf a 0m =F Ayer em 7 On) (by caution Gem “ay—T* This lst formula looks somewhat messy until we realize that advantage of small amounts of sloppiness and use an infinite decreasing geometric ies as an upper bound. Backing up one step and applying equation (A.6), we have ‘an again take 3) Ta = (=) en? + O(n'=3) 3 (2) en? Ines3 < Lz) en + (nh) - 1 une = oe on) 16 2 nie? = Fer oui) = O(n) ‘Thus, we have derived a guess of T(n) = O(n?) for our original recurrence T(n) = 31 (|n/4)) + O(n). In this example, the coefficients of en? form a decreasing geometric series and, by equation (A.6), the sum of these coefficients4A The recursion-tree method for solving recurrences or oN : ea) oF / \ / \ ee) Ge Total: O(n Figure 4.6 A recursion tee forthe recurrence Tan) = T(n/3) + TQn/3) en. is bounded from above by the constant 16/13. Since the root's contribution to the total cost is cn?, the root contributes a constant fraction of the total cost. In other words, the cost of the root dominates the total cost of the tree. In fact, if O(n2) is indeed an upper bound for the recurrence (as we shall verify in momen, then it must be a tight bound. Why? The first I contributes a cost of @(n7), and so ©)(7;2) must be a lower bound for the recurrence. Now we ean use the substitution method to verify that our guess was cor ‘rect, that is, T(m) = O(n?) is an upper bound for the recurrence T(n) = “1 (0). We want to show that T(n) = dn? for some constant d > 0, Using the same constant ¢ > 0 as before, we have Tin) = 37(\n/4)) +en? 3d [n/4)? + en? = 3d(n/4)? + en? 3 = Sdn? en 16 = dn?, where the last step holds as long as d > (16/13)c In another, more intricate, example, Figure 4.6 shows the recursion tree for T(n) =T(n/3) + TQn/3) ~ O(n) (Again, we omit floor and ceiling functions for simplicity.) As before, we let ¢ represent the constant factor in the O(7) term. When we add the values across the levels of the recursion tree shown in the figure, we get a value of en for every level.2 Chaprer 4 Divide-and-Conguer The longest simple path from the root to a leaf ism + (2/3) > (2/3)'m » 1. Since (2/3)kn = 1 when k = logsy2., the height of the tree is logs, Intuitively, we expect the solution to the recurrence to be at most the number of levels times the cost of each level, or O(cn logan) = O(n ign). Figure 4.6 shows only the top levels of the recursion tree, however, and not every level in the tree contributes a cost of ext. Consider the cost of the leaves. If this recursion tree were a complete binary tree of height logaj27, there would be 2659/2" = n'e/2? leaves. Since the cost of each leaf is a constant, the total cost of all leaves would then be @(n'*327) which, since log3/22 is a constant strictly greater than 1, is w(n Ign). This recursion tree is not a complete binary tree, however, and so it has fewer than n=? leaves. Moreover, as we go down from the root, more and more internal nodes are absent. Consequently, levels toward the bottom of the recursion tree contribute less than cv to the total cost. We could work out an accu rate accounting of all costs, but remember that we are just trying to come up with a guess to use in the substitution method. Let us tolerate the sloppiness and attempt to show that a guess of O(n Ig) for the upper bound is correct. Indeed, we can use the substitution method to verify that O(r Ig.) is an upper bound for the solution to the recurrence. We show that T(n) < dn ign, where d is a suitable positive constant. We have Tin) = T(n/3) 4 TQn/3) -en = d(n/3)1g(x/3) + d2n/3) 1g(2n/3) + en = (d(n/3)lgn —d(n/3) 1g 3) + (dQn/3) 1g — d(2n/3)1g(3/2)) 4 en = dnign —d((n/3)1g3 + 2n/3)1g(3/2)) ~ en dn ign —d((n/3)1g3 | (2n/3)1g3 —2n/3)1g2) en = dnign—dn(lg3—2/3) +n dnign, aslongas d » accounting o! ¢/(lg3—(2/3)). Thus, we did not need to perform a more accurate ts in the recursion tree. Exercises. 44-1 Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = 3T (\n/2|) 4 n. Use the substitution method to verify your answer, 442 Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = T(n/2) +12. Use the substitution method to verily your answer.45° The master method for solving recurrences 93 43 Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) =4T(n/2 4 2) 4 n. Use the substitution method to verify your answer. 444 Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = 21 (n = 1) + L. Use the substitution method to verify your answer. Use a recursion tree to determine a good asymptotic upper bound on the recurrence T(n) = T(n=1) +7 (n/2)-+n. Use the substitution method to verify your answer. 46 Argue that the solution to the recurrence T(n) = T(n/3) +T(Qn/3)-+en, where c is a constant, is (0 lg) by appealing to a recursion tree. 447 Draw the recursion tree for T(n) = 47(|n/2|) + en, where ¢ is a constant, and provide a tight asymptotic bound on its solution. Verify your bound by the substi- tution method. 448 Use a recursion tree to give an asymptotically tight solution to the recurrence T(n) =T(n —a) + T(a) + en, where a > 1 and ¢ > 0 are constants 449 Use a recursion tree to give an asymptotically tight solution to the recurrence T(n) = Tan) ~T (0 —ain)-+ en, where «isa constant in the range () = @ = | and ¢ > Os also a constant. 4.5. The master method for solving recurrences ‘The master method provides a “cookbook” method for solving recurrences of the form T(n) =aT(n/b) + fin). (4.20) where @ > 1 and b > 1 are constants and f(n) is an asymptotically positive function. To use the master method, you will need to memorize th . but then you will be able to solve many recurrences quite easily, often without pencil and paper.