Scribd
Upload
2 views4 pages

Advanced Algorithms Lecture 09.0

The document discusses the Min Cost Circulation Problem in directed graphs, defining key concepts such as circulation, cost, and residual graphs. It highlights special cases like Max-Flow and Shortest Path problems, and introduces the notion of strongly polynomial algorithms. The document also outlines an algorithm for finding a min-cost circulation by constructing a residual graph and checking for negative cost cycles.

Uploaded by

mhmad240
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
2 views4 pages

Advanced Algorithms Lecture 09.0

The document discusses the Min Cost Circulation Problem in directed graphs, defining key concepts such as circulation, cost, and residual graphs. It highlights special cases like Max-Flow and Shortest Path problems, and introduces the notion of strongly polynomial algorithms. The document also outlines an algorithm for finding a min-cost circulation by constructing a residual graph and checking for negative cost cycles.

Uploaded by

mhmad240
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

MIT OpenCourseWare

http://ocw.mit.edu

6.854J / 18.415J Advanced Algorithms


Fall 2008
��

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
18.415/6.854 Advanced Algorithms October 15, 2001

Lecture 9
Lecturer: Michel X. Goemans Scribe: Jasrper Lin

1 Min Cost Circulation Problem


Given a directed graph G = (V,E ) , an anti-symmetric1 cost function c : E + Z, and a capacity
function u : E + Z.

Definition 1 A pow is a mapping f : E +R that is anti-symmetric1 and f (v, w) 5 u(v, w),


V(V,W)E E.

Definition 2 The cost of a pow f is defined to be

Definition 3 A circulation is a pow f such that

The min cost circulation problem is to find a circulation of minimum cost.

1.1 Special Cases of Min Cost Circulation Problem


The min cost circulation problem is actually fairly general. In includes the following important
problems as special cases:

Max-Flow
Min Cost Flow
Min Weighted Bipartite Matching
a Shortest Path
Given a directed graph G = (V,E ) , a function I : E + R that specifies the lengths of the
edges, and a fixed vertex s E V, find the shortest simple2 path from s to t , for every vertex
t E V. I f the length of edges are nonnegative numbers, then Dijkstra's Algorithm may be
used. Dijkstra's Algorithm has different running times depending on how the priority queue
+
is implemented. Using a Fibonacci Heap, the running time is O(m n log n), where m = IEl
and n = IVI. In general, if 1 can be negative, the problem is NP-complete. However, it is
common to consider the case when there is no negative length cycles. Then the Bellman-Ford
Algorithm can be used. This algorithm either detects that there is a negative length cycle or
computes, in O(mn), the length of the shortest path between s and t for every vertex t E V.
I A function f (x,y) is anti-symmetric iff f (x,y) = -f (y, x).
2A path is simple if it visit each vertex at most once.
Remark 1 The algorithms mentioned here do not i n general apply directly t o undirected graphs as
one would think. The naive approach would be to turn every edge i n the undirected graph into two
edges going both ways i n the digraph. This approach works if the lengths of the edges are nonnegative
numbers. However, if an undirected graph has an edge with a negative length, doubling it would create
a negative length cycle where one might not have existed i n the first place. A s it turns out, i n this
case, the problem i n undirected graphs is much more dificult than the digraph shortest path problem
and some underlying assumptions that we make about digraphs do not hold for undirected graphs.
There is a n algorithm for this problem using non-bipartite matching algorithm.

1.2 Strongly Polynomial Running Time


Definition 4 The running time of an algorithm is strongly polynomial iff

the number of operations required is bounded by a polynomial of the number of data items i n
the input, and not the size of the numerical values i n the input.
all arithmetic operations are performed on numbers whose size is polynomial i n input size.

If the running time of an algorithm is polynomial but not strongly polynomial, then we say that it is
weakly polynomial.

Our goal is to search for a strongly polynomial algorithm to solve the min cost circulation problem.
Not all polynomial time algorithms are strongly polynomial. The algorithms for Linear Programming
and some versions of Dijkstra's Algorithm are not strongly polynomial in running time. For example,
+
using different priority queues, both O(n log log L) and O(m n d w ) are possible, where L is the
maximum length in the graph.

1.3 Searching for an Algorithm


Our first goal is to devise a check that given a flow f determines whether or not f is optimal.

Definition 5 Given a graph G = (V,E ) and a circulation f , we define the residual graph o f f to be
Gf = (V,Ef ), where Ef = {(u,W) E Elf(v, w) < u(u, w)), and the capacity function is replaced by
the residual capacity function uf (u,u) = u(u, u) - f (u,u).

The important property of the residual graph is that if g is a circulation in the residual graph G f ,
+
then f g is a valid circulation in G, since f + +
g 5 f u - f = u.
A vertex potential p is a real-valued function defined on the set of vertices (p : V + Kt). Corre-
+
sponding to a potential p, we can define a reduced cost function cp(u,w) = c(u,w) p(u) - p(w).

Lemma I For every directed cycle C, cp(C) = c(C), where the notation cp(C) = Cc cp(u,w).

Proof of Lemma 1: cp(C) = Cc c,(v, w) = C,(c(v, w) +p(v) -p(w)) = CCc(v, w )+C,(p(u) -


p(w)) = c(C), since every u E C appears again as a w E C the same amount of times since C is a
cycle.

Lemma 2 For every circulation f and potential p,


Proof of Lemma 2: One way is to show that a circulation can be decomposed into a sum of
many cycles, and then use Lemma 1. Here we give a direct proof:

Theorem 3 (Optimality Conditions) Let f be a circulation. The following are equivalent:

(i) f is a min-cost circulation


(ii) Gf has no directed cycles of negative cost
(iii) There is a potential p such that for every (u, w) E Ef, cp(u, w) > 0.
Proof of Theorem 3:

( l i + l i i i ) Assume there is a circulation f* with c . f * < c . f and a potential p with cp(u,w) >0
for every (u,w) E Ef.By Lemma 2, c, - f* < cp . f . Then

However, (u, w) E Ef implies that f * (v,w) > f (u, w). Therefore, 2 C(v,w)
cP(v,w) .
(f* (u, w) - f (u, w)) 2 0, which is a contradiction.
(i + ii) Suppose f is of min-cost and there exists a cycle C in Gfsuch that c(C) < 0. Form f I, where
+
fl(u, u) = f (u, u) E when (u, v) E C, fl(u, v) = f (u, u) - c when (v, w) E C , f t ( u , u) = f (u, u)
otherwise, and E = minetC cf ( e ) > 0. Therefore, c(f I) = c(f )+2~.c(C)< c(f ) , a contradiction.
(ii -+iii) Will be proved next lecture.

Remark 2 W e now have an algorithm for finding a min-cost circulation:

1. Construct the residual graph.


2. Check if there is any negative cost cycles. If none, end.
3. If one is found, push flow and repeat.

Step 3 can be done i n many diflerent ways. They key is to choose intelligently to reduce the number
of iterations necessary.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy