lecture 5

36
Lecture 5 Today, how to solve recurrences We learned “guess and proved by induction” We also learned “substitution” method Today, we learn the “master theorem” More divide and conquer: closest pair problem matrix multiplication

Upload: palani

Post on 04-Jan-2016

18 views

Category:

Documents


0 download

DESCRIPTION

Lecture 5. Today, how to solve recurrences We learned “guess and proved by induction” We also learned “substitution” method Today, we learn the “master theorem” More divide and conquer: closest pair problem matrix multiplication. Master Theorem. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Lecture 5

Lecture 5

Today, how to solve recurrences We learned “guess and proved by induction” We also learned “substitution” method Today, we learn the “master theorem”

More divide and conquer: closest pair problem matrix multiplication

Page 2: Lecture 5

Master Theorem

Theorem 4.1 (CLRS, Theorem 4.1) Let a ≥ 1 and b > 1 be constants. Let f(n) be a function and let T(n) be defined on the nonnegative integers by

T(n) = aT(n/b) + f(n).

Then

1

0

largefor )()/(

AND )(

)(

)(

)(

log)(

log

log

log

log

log

c

nncfbnaf

nnf

nnf

nOnf

nf

nn

n

nT

a

a

a

a

a

b

b

b

b

b

Page 3: Lecture 5

Note

Only apply to a particular family of recurrences.

f(n) is positive for large n. Key is to compare f(n) with nlog_b a Case 2, more general is f(n) = Θ( nlog_b a lgkn).

Then the result is T(n) = Θ( nlog_b a lgk+1n). Sometimes it does not apply. Ex. T(n) =

4T(n/2) + n2 /logn.

Page 4: Lecture 5

Proof ideas of Master Theorem Consider a tree with T(n) at the root, and apply the recursion to

each node, until we get down to T(1) at the leaves. The first recursion is T(n) = aT(n/b) + f(n), so assign a cost of f(n) to the root. At the next level we have “a” nodes, each with a cost of T(n/b). When we apply the recursion again, we get a cost of af(n/b) for all of these. At the next level we have a2 nodes, each with a cost of T(n/b2). We get a cost of a2f(n/b2). We continue down to T(1) at the leaves. There are alog_b n leaves and each costs Θ(1), which gives Θ(alog_b n). The total cost associated with f is Σ 0 ≤ i ≤ log_b n - 1 ai f(n/bi).

Thus T(n) = Θ(n log_b a) + Σ 0 ≤ i ≤ (log_b n) - 1 ai f(n/bi). The three cases now come from deciding which term is

dominant. In case (1), the Θ term is dominant. In case (2), the terms are roughly equal (but the second term has an extra lg n factor). In case (3), the f(n) term is dominant. The details are somewhat painful, but can be found in CLRS, pp. 76-84.

Page 5: Lecture 5

f (n/b)

Idea of master theorem

f (n/b) f (n/b)

(1)

…Recursion tree:

f (n) a

f (n/b2)f (n/b2) f (n/b2)…ah = logbn

f (n)

a f (n/b)

a2 f (n/b2)

#leaves = ah

= alogbn

= nlogba

nlogba(1)

Page 6: Lecture 5

Three common cases

Compare f (n) with nlogba:1. f (n) = O(nlogba – ) for some constant > 0.

• f (n) grows polynomially slower than nlogba (by an n factor).

Solution: T(n) = (nlogba) .

Page 7: Lecture 5

f (n/b)

Idea of master theorem

f (n/b) f (n/b)

(1)

…Recursion tree:

f (n) a

f (n/b2)f (n/b2) f (n/b2)…ah = logbn

f (n)

a f (n/b)

a2 f (n/b2)

nlogba(1)CASE 1: The weight increases geometrically from the root to the leaves. The leaves hold a constant fraction of the total weight.

CASE 1: The weight increases geometrically from the root to the leaves. The leaves hold a constant fraction of the total weight. (nlogba)

These functions increasefrom top to bottomgeometrically, hence weonly need to have thelast bottom term

Page 8: Lecture 5

Case 2

Compare f (n) with nlogba:

2. f (n) = (nlogba lgkn) for some constant k 0.

• f (n) and nlogba grow at similar rates.

• This is clear for k=0. For k>0, the intuition is that lgk n factor remain for constant fraction of levels, hence sum to the following

Solution: T(n) = (nlogba lgk+1n) .

Page 9: Lecture 5

f (n/b)

Idea of master theorem

f (n/b) f (n/b)

(1)

…Recursion tree:

f (n) a

f (n/b2)f (n/b2) f (n/b2)…ah = logbn

f (n)

a f (n/b)

a2 f (n/b2)

nlogba(1)CASE 2: (k = 0) The weight is approximately the same on each of the logbn levels.

CASE 2: (k = 0) The weight is approximately the same on each of the logbn levels.

(nlogbalg n)

All levels same

Page 10: Lecture 5

Case 3, c<1, akf(n/bk) geometrically decreases hence = Θ(f(n))

Compare f (n) with nlogba:

3. f (n) = (nlogba + ) for some constant > 0.

• f (n) grows polynomially faster than nlogba (by an n factor),

and f (n) satisfies the regularity condition that a f (n/b) c f (n) for some constant c < 1.

Solution: T(n) = ( f (n) ) .

Page 11: Lecture 5

f (n/b)

Idea of master theorem

f (n/b) f (n/b)

(1)

…Recursion tree:

f (n) a

f (n/b2)f (n/b2) f (n/b2)…ah = logbn

f (n)

a f (n/b)

a2 f (n/b2)

nlogba(1)CASE 3: The weight decreases geometrically from the root to the leaves. The root holds a constant fraction of the total weight.

CASE 3: The weight decreases geometrically from the root to the leaves. The root holds a constant fraction of the total weight. ( f (n))

af(n/b)<(1-ε)f(n)

Page 12: Lecture 5

Examples for the Master Theorem

The Karatsuba recurrence has a = 3, b = 2, f(n) = cn. Then case 1 applies, and so T(n) = Θ(n l og_2 3 ), as we found.

The mergesort recurrence has a = 2, b = 2, f(n) = n. Then case 2 applies, and so T(n) = Θ(n lg n).

Finally, a recurrence like T(n) = 3T(n/2) + n2 gives rise to case 3. In this case f(n) = n2, so 3f(n/2) = 3 (n/2)2 = (3/4) n2 ≤ c n2 for c = 3/4, and so T(n) = Θ(n2).

Note that the master theorem does not cover all cases. In particular, it does not cover the case

T(n) = 2 T(n/2) + n / lg n since then the only applicable case is case 3, but then the

inequality involving f does not hold.

Page 13: Lecture 5

Closest pair problem

Input: A set of points P = {p1,…, pn} in two

dimensions Output:

The pair of points pi, pj that minimize the Euclidean distance between them.

Page 14: Lecture 5

Distances

Euclidean distance

1x 2x

2y

1y 11, yx

22 , yx

221

2212211 ,, yyxxyxyx

Page 15: Lecture 5

Closest Pair Problem

Page 16: Lecture 5

Closest Pair Problem

Page 17: Lecture 5

Divide and Conquer

O(n2) time algorithm is easy Assumptions:

No two points have the same x-coordinates No two points have the same y-coordinates

How do we solve this problem in 1 dimension? Sort the number and walk from left to right to find

minimum gap.

Page 18: Lecture 5

Divide and Conquer

Divide and conquer has a chance to do better than O(n2).

We can first sort the points by their x-coordinates and sort also by y-coordinates

Page 19: Lecture 5

Closest Pair Problem

Page 20: Lecture 5

Divide and Conquer for the Closest Pair Problem

Divide by x-median

Page 21: Lecture 5

Divide

Divide by x-median

L R

Page 22: Lecture 5

Conquer

Conquer: Recursively solve L and R

L R

1

2

Page 23: Lecture 5

Combination I

Take the smaller one of 1 , 2 : = min(1 , 2 )

L R

2

Page 24: Lecture 5

Combination IIIs there a point in L and a point in R whose distance is smaller than ?

= min(1 , 2 )

L R

Page 25: Lecture 5

Combination II

If the answer is “no” then we are done!!! If the answer is “yes” then the closest such

pair forms the closest pair for the entire set

How do we determine this?

Page 26: Lecture 5

Combination IIIs there a point in L and a point in R whose distance is smaller than ?

L R

Page 27: Lecture 5

Combination IIIs there a point in L and a point in R whose distance is smaller than ?

Need only to consider the narrow bandO(n) time

L R

Page 28: Lecture 5

Combination IIIs there a point in L and a point in R whose distance is smaller than ?

Denote this set by S, assume Sy is the sorted list of S by the y-coordinates.

L R

Page 29: Lecture 5

Combination II

There exists a point in L and a point in R whose distance is less than if and only if there exist two points in S whose distance is less than .

If S is the whole thing, did we gain anything?

CLAIM: If s and t in S have the property that ||s-t|| < , then s and t are within 15 positions of each other in the sorted list Sy.

Page 30: Lecture 5

Combination IIIs there a point in L and a point in R whose distance is smaller than ?

L R

There are at most one point in each box of size δ/2 by δ/2. Thus s and t cannot be too far apart.

Page 31: Lecture 5

Closest-Pair Preprocessing:

Construct Px and Py as sorted-list by x- and y-coordinates

Closest-pair(P, Px,Py) Divide

Construct L, Lx , Ly and R, Rx , Ry Conquer

Let 1= Closest-Pair(L, Lx , Ly ) Let 2= Closest-Pair(R, Rx , Ry )

Combination Let = min(1 , 2 ) Construct S and Sy For each point in Sy, check each of its next 15 points down

the list If the distance is less than , update the as this smaller

distance

Page 32: Lecture 5

Complexity Analysis

Preprocessing takes O(n lg n) time Divide takes O(n) time Conquer takes 2 T(n/2) time Combination takes O(n) time

T(n) = 2T(n/2) + cn

So totally takes O(n lg n) time.

Page 33: Lecture 5

Matrix Multiplication

Suppose we multiply two NxN matrices together. Regular method is NxNxN = N3 multiplications O(N3)

Page 34: Lecture 5

Can we Divide and Conquer?

C11 = A11*B11 + A12*B21 C12 = A11*B12 + A12*B22 C21 = A21*B11 + A22*B21

C22 = A21*B12 + A22*B22

Complexity : T(N) = 8T(N/2) + O(N2) = O(Nlog

28) = O(N3)

2221

1211

AA

AAA =

2221

1211

BB

BB B =

2221

1211

CC

CC C= A*B =

No improvement

Page 35: Lecture 5

Strassen’s Matrix Multiplication

P1 = (A11+ A22)(B11+B22)

P2 = (A21 + A22) * B11

P3 = A11 * (B12 - B22)

P4 = A22 * (B21 - B11)

P5 = (A11 + A12) * B22

P6 = (A21 - A11) * (B11 + B12)

P7 = (A12 - A22) * (B21 + B22)

C11 = P1 + P4 - P5 + P7

C12 = P3 + P5

C21 = P2 + P4

C22 = P1 + P3 - P2 + P6

And do this recursively as usual.

VolkerStrassen

Page 36: Lecture 5

Time analysis

T(n) = 7T(n/2) + O(n2 )

= 7logn by the Master Theorem

=nlog7

=n2.81

Best bound: O(n2.376) by Coppersmith-Winograd.

Best known (trivial) lower bound: Ω(n2). Open: what is the true complexity of matrix

multiplication?