Scribd
Upload
39 views38 pages

Daniel Kane: Advanced Algorithms and Complexity Data Structures and Algorithms

The document summarizes the Edmonds-Karp algorithm for finding the maximum flow in a network. It uses the Ford-Fulkerson algorithm but always chooses the shortest augmenting path using BFS. This guarantees that no edge is saturated more than V times, leading to a runtime of O(V*E^2). The algorithm improves on Ford-Fulkerson by removing dependence on numerical capacities.

Uploaded by

sbei arafet
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
39 views38 pages

Daniel Kane: Advanced Algorithms and Complexity Data Structures and Algorithms

The document summarizes the Edmonds-Karp algorithm for finding the maximum flow in a network. It uses the Ford-Fulkerson algorithm but always chooses the shortest augmenting path using BFS. This guarantees that no edge is saturated more than V times, leading to a runtime of O(V*E^2). The algorithm improves on Ford-Fulkerson by removing dependence on numerical capacities.

Uploaded by

sbei arafet
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/ 38

Flows in Networks:

The Edmonds-Karp
Algorithm

Daniel Kane
Department of Computer Science and Engineering
University of California, San Diego

Advanced Algorithms and Complexity


Data Structures and Algorithms
Learning Objectives
Implement the Edmonds-Karp algorithm.
Understand the runtime bound.
Last Time

Ford-Fulkerson algorithm for Maxflow.


Runtime O(|E ||f |).
Sometimes very slow if graph has large
capacities.
Bad Example
Bad Example
Bad Example
Bad Example
Bad Example
Bad Example
Bad Example
Bad Example
Bad Example
Fix

Fortunately, the Ford-Fulkerson algorithm


gives us a choice as to which augmenting
path to use.
Fix

Fortunately, the Ford-Fulkerson algorithm


gives us a choice as to which augmenting
path to use.
Is there a way to choose augmenting paths
to avoid this kind of runtime problem?
Edmonds-Karp Algorithm

Use the Ford-Fulkerson algorithm, always


choosing the shortest (in terms of number of
edges) augmenting path.
Edmonds-Karp Algorithm

Use the Ford-Fulkerson algorithm, always


choosing the shortest (in terms of number of
edges) augmenting path.
Find augmenting paths using BFS instead of
DFS.
Edmonds-Karp Execution
Edmonds-Karp Execution
Edmonds-Karp Execution
Edmonds-Karp Execution
Edmonds-Karp Execution
Augmenting Paths
Need to analyze augmenting paths.
.
Augmenting Paths
Augmenting flow always saturates (uses all
the available flow from) an edge.
Augmenting Paths
Changes to residual network.
.
Analysis Idea

Show that no edge is saturated too


many times.
Fails to hold in the bad case, where the
middle edge is repeatedly saturated.
Increasing Distances
We will need this critical Lemma:
Lemma
As the Edmonds-Karp algorithm executes,
for any vertex v ∈ V the distance dGf (s, v )
only increases.
Similarly, dGf (v , t) and dGf (s, t) can only
increase.
Proof
New edges all point backwards along
augmenting path.
Proof
Augmenting path is a shortest path.
.
Proof
All new edges point from vertices
further from s to vertices closer.
Proof
New edges do not help you get from s
to v faster.
Reuse Limit

Lemma
When running the Edmonds-Karp algorithm,
if an edge e is saturated, it will not be used
in an augmenting path again, until dGf (s, t)
increases.
Proof
Initially, d (s, u) = x, d (v , t) =
y , d (s, t) = x + y + 1.
When used again, d (s, v ) ≥ x + 1,
d (u, t) ≥ y + 1.
Therefore, when used again,
d (s, t) ≥ (x +1)+(y +1)+1 = x +y +3.
Analysis

dGf (s, t) can only increase |V | times.


Each time can only have O(|E |) many
saturated edges.
Therefore, only O(|V ||E |) many
augmenting path.
Each path takes O(|E |) time.
Total runtime: O(|V ||E |2).
Problem
Which of the following is true about the
Edmonds-Karp algorithm:
1 No edge is saturated more than |V |

times.
2 The lengths of the augmenting paths

decrease as the algorithm progresses.


3 Changing the capacities of edges will

not affect the runtime.


Solution
Which of the following is true about the
Edmonds-Karp algorithm:
1 No edge is saturated more than |V |

times.
2 The lengths of the augmenting paths

decrease as the algorithm progresses.


3 Changing the capacities of edges will

not affect the runtime.


Summary

Choose augmenting paths based on path


length.
Removes runtime dependence on
numerical sizes of capacities.
Summary

Choose augmenting paths based on path


length.
Removes runtime dependence on
numerical sizes of capacities.
There are better, much more
complicated algorithms.
State of the art O(|V ||E |).
Next Time

Applications of Maxflow algorithms.

You might also like