 ## Slide #1.

The Design and Analysis of Algorithms Chapter 10: Iterative Improvement The Maximum Flow Problem
More slides like this ## Slide #2.

The Maximum Flow Problem    Introduction Definition of a Flow The Ford-Fulkerson Method  Residual networks  Augmenting Paths  Shortest-Augmenting-Path Algorithm  Definition of a Cut  Max-Flow Min-Cut Theorem  Time efficiency 2
More slides like this ## Slide #3.

Introduction A flow network is a model of a system where some material is produced at its source, travels through the system, and is consumed at its end.      Connected weighted digraph with n vertices One vertex with no entering edges, called the source one vertex with no leaving edges, called the sink edge capacity: positive integer weight uij on each directed edge (i.j) The maximum flow problem is the problem of maximizing the flow of a material through the network without violating any capacity constraints 3
More slides like this ## Slide #5.

Definition of a Flow A flow is an assignment of real numbers xij to edges (i,j) of a given network that satisfies the following constraints:  flow-conservation requirements  (j,i) E xji = (i,k) E xik  capacity constraints 0 ≤ xij ≤ uij for every edge (i,j)  E  flow value ∑ x1j = ∑ xkn 5
More slides like this ## Slide #6.

The Ford-Fulkerson Method  At each iteration it increases the flow  Based on:  Residual networks  Augmenting paths  Computes the minimum Cut 6
More slides like this ## Slide #7.

Residual networks Given a flow network and a flow, the residual network consists of those edges that can take more flow. if uij – xij > 0, (i,j) is in the residual network 7
More slides like this ## Slide #8.

Augmenting Paths  An augmenting path is a simple path in the underlying un-oriented graph of the residual network from the source s to the sink t.  Each edge in the augmenting path admits some additional positive flow.  The flow that can pass along the augmenting path is used to increase the flow through the original network 8
More slides like this ## Slide #9.

Augmenting Path (Ford-Fulkerson) Method  Start with the zero flow (xij = 0 for every edge)  On each iteration, try to find a flow-augmenting path from source to sink, along which some additional flow can be sent  If a flow-augmenting path is found, adjust the flow along the edges of this path to get a flow of increased value and try again  If no flow-augmenting path is found, the current flow is maximum 9
More slides like this ## Slide #10.

Finding a flow-augmenting path  Consider paths from source to sink in which any two consecutive vertices i,j are either  forward edge ( → ) from i to j with some positive unused capacity : rij = uij – xij OR  backward  edge ( ← ) from j to i with positive flow xji If a flow-augmenting path is found, the current flow can be increased by r units by increasing xij by r on each forward edge and decreasing xji by r on each backward edge, where r = min {rij on all forward edges, xji on all backward edges} 10
More slides like this ## Slide #11.

Example 5 0/3 0/2 1 2 0/3 xij/uij 0/5 0/4 3 0/2 6 0/1 4 Augmenting path: 1→2 →3 →6 11
More slides like this ## Slide #12.

Example 5 0/3 2/2 1 2 0/3 2/5 0/4 3 2/2 6 0/1 4 Augmenting path: 1 →4 →3←2 →5 →6 12
More slides like this ## Slide #14.

Specifics of the method  Assuming the edge capacities are integers, r is a positive integer  Maximum value is bounded by the sum of the capacities of the edges leaving the source; hence the augmenting-path method has to stop after a finite number of iterations  The final flow is always maximum, its value doesn’t depend on a sequence of augmenting paths used  The augmenting-path method doesn’t prescribe a specific way for generating flow-augmenting paths  Selecting a bad sequence of augmenting paths could impact the method’s efficiency 14
More slides like this ## Slide #15.

Shortest-Augmenting-Path Algorithm Starting at the source, perform BFS traversal by marking new (unlabeled) vertices with two labels:  first label – indicates the amount of additional flow that can be brought from the source to the vertex being labeled  second label – indicates the vertex from which the vertex being labeled was reached, with “+” or “–” added to the second label to indicate whether the vertex was reached via a forward or backward edge 15
More slides like this ## Slide #16.

Vertex Labeling  The source is always labeled with ∞,-  All other vertices are labeled as follows:  If unlabeled vertex j is connected to the dequeued vertex i of the traversal queue by a directed edge from i to j with positive unused capacity rij where lj  = uij –xij (forward edge), vertex j is labeled with lj,i+, = min{li, rij} If unlabeled vertex j is connected to the dequeued vertex i of the traversal queue by a directed edge from j to i with positive flow xji (backward edge), vertex j is labeled lj,i-, where lj = min{li, xji} 16
More slides like this ## Slide #17.

Vertex Labeling  If the sink ends up being labeled, the current flow can be augmented by the amount indicated by the sink’s first label  The augmentation of the current flow is performed along the augmenting path traced by following the vertex second labels from sink to source; the current flow quantities are increased on the forward edges and decreased on the backward edges of this path  If the sink remains unlabeled after the traversal queue becomes empty, the algorithm returns the current flow as maximum and stops 17
More slides like this ## Slide #18.

Example 5 0/3 0/2 1 2 0/4 0/5 3 0/2 6 2,2+ 0/3 0/1 5 0/3 4 ∞,- Queue: 1 2 4 3 5 6 ↑↑↑↑ 0/2 1 2,1+ 0/4 0/5 2 2,3+ 0/2 3 6 2,2+ 0/3 0/1 4 3,1+ Augment the flow by 2 (the sink’s first label) along the path 1→2→3→6 18
More slides like this ## Slide #19.

Example 5 0/3 2/2 1 2 0/4 2/5 3 2/2 6 1,2+ 0/3 0/1 5 0/3 4 ∞, - 1 2/2 0/3 Queue: 1 4 3 2 5 6 ↑↑↑↑↑ 1,3- 2 0/4 2/5 3 2/2 1,4+ 1,5+ 6 0/1 4 3,1+ Augment the flow by 1 (the sink’s first label) along the path 1→4→3←2→5→6 19
More slides like this ## Slide #20.

Example 5 1/3 2/2 1 1/4 1/5 2 1/3 3 2/2 6 1/1 5 1/3 4 Queue: 1 4 ↑↑ ∞,- 2/2 1 2 1/4 1/5 1/3 3 2/2 6 1/1 4 2,1+ No augmenting path (the sink is unlabeled) the current flow is maximum 20
More slides like this ## Slide #22.

Definition of a Cut X - a set of vertices in a network that includes its source but does not include its sink  Xc - the complement of X, including the sink   C(X,Xc) : the cut induced by this partition of the vertices - the set of all the edges with a tail in X and a head in Xc.  c(X,Xc) : Capacity of a cut - the sum of capacities of the edges that compose the cut.  Note that if all the edges of a cut were deleted from the network, there would be no directed path from source to sink  Minimum cut is a cut of the smallest capacity in a given network 22
More slides like this ## Slide #23.

Examples of network cuts 5 4 3 2 1 2 3 5 3 2 6 1 4 If X = {1} and X = {2,3,4,5,6}, C(X,Xc) = {(1,2), (1,4)}, c=5 If X = {1,2,3,4,5} and X = {6}, C(X,Xc) = {(3,6), (5,6)}, c=6 If X = {1,2,4} and X ={3,5,6}, C(X,Xc) = {(2,3), (2,5), (4,3)}, c=9 23
More slides like this ## Slide #24.

Max-Flow Min-Cut Theorem  The value of maximum flow in a network is equal to the capacity of its minimum cut  The shortest augmenting path algorithm yields both a maximum flow and a minimum cut:  maximum flow is the final flow produced by the algorithm  minimum cut is formed by all the edges from the labeled vertices to unlabeled vertices on the last iteration of the algorithm  all the edges from the labeled to unlabeled vertices are full, i.e., their flow amounts are equal to the edge capacities, while all the edges from the unlabeled to labeled vertices, if any, have zero flow amounts on them 24
More slides like this ## Slide #25.

Time Efficiency  The number of augmenting paths needed by the shortestaugmenting-path algorithm never exceeds nm/2, where n and m are the number of vertices and edges, respectively  Since the time required to find shortest augmenting path by breadth-first search is in O(n+m)=O(m) for networks represented by their adjacency lists, the time efficiency of the shortest-augmenting-path algorithm is in O(nm2) for this representation  More efficient algorithms have been found that can run in close to O(nm) time, but these algorithms don’t fall into the iterative-improvement paradigm 25
More slides like this