328 lines
17 KiB
HTML
328 lines
17 KiB
HTML
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
|
|
<html xmlns="http://www.w3.org/1999/xhtml" xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office">
|
|
|
|
<head>
|
|
<meta http-equiv="Content-Language" content="en-us" />
|
|
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
|
|
<title>Strongly Connected Components - Lecture by Rashid Bin Muhammad, PhD.</title>
|
|
<meta name="Author" content="Rashid Bin Muhammad, PhD."/>
|
|
<style type="text/css">
|
|
.style1 {
|
|
font-size: large;
|
|
}
|
|
.style2 {
|
|
font-size: large;
|
|
margin-left: 80px;
|
|
}
|
|
.style3 {
|
|
font-size: large;
|
|
text-align: left;
|
|
}
|
|
.style4 {
|
|
text-decoration: underline;
|
|
}
|
|
.style5 {
|
|
font-size: large;
|
|
text-align: center;
|
|
}
|
|
.style6 {
|
|
margin-left: 80px;
|
|
}
|
|
.style7 {
|
|
font-size: large;
|
|
color: #800000;
|
|
}
|
|
.style8 {
|
|
color: #0000FF;
|
|
}
|
|
.style9 {
|
|
font-size: large;
|
|
color: #0000FF;
|
|
}
|
|
.style10 {
|
|
text-align: center;
|
|
}
|
|
.style11 {
|
|
font-size: large;
|
|
margin-left: 40px;
|
|
}
|
|
.style12 {
|
|
text-align: right;
|
|
font-family: "Blackadder ITC";
|
|
}
|
|
.style13 {
|
|
font-family: "Blackadder ITC";
|
|
font-size: xx-large;
|
|
color: #FF0000;
|
|
}
|
|
.style14 {
|
|
text-align: left;
|
|
}
|
|
.style15 {
|
|
font-size: x-large;
|
|
}
|
|
.style16 {
|
|
font-family: Symbol;
|
|
}
|
|
.style17 {
|
|
font-family: "Times New Roman";
|
|
}
|
|
</style>
|
|
</head>
|
|
|
|
<body background="../../../Maingif/Bck2.gif" link="#0000FF" vlink="#0000FF" alink="#FF0000" BODYLINK="blue">
|
|
|
|
<p class="style5"><font size="4"><img SRC="../../../Maingif/redline.gif" height=2 width=640/></font></p>
|
|
<h1 class="style10">Strongly Connected Components</h1>
|
|
<p class="style5"><font size="4"><img SRC="../../../Maingif/redline.gif" height=2 width=640/></font></p>
|
|
<p class="style1"><span class="style13">D</span>ecomposing a directed graph into its strongly connected
|
|
components is a classic application of depth-first search. The problem of
|
|
finding connected components is at the heart of many graph application.
|
|
Generally speaking, the connected components of the graph correspond to
|
|
different classes of objects. The first linear-time algorithm for strongly
|
|
connected components is due to Tarjan (1972). Perhaps, the algorithm in the CLRS
|
|
is easiest to code (program) to find strongly connected components and is due to
|
|
Sharir and Kosaraju.</p>
|
|
<p class="style1">Given digraph or
|
|
directed graph G = (V, E), a strongly connected component (SCC) of G is a
|
|
maximal set of vertices C subset of V, such that for all <em>u</em>, <em>v</em>
|
|
in C, both
|
|
<em>u</em> <span class="style16">Þ</span> <em>v</em> and <em>v</em>
|
|
<span class="style16">Þ</span> <em>u</em>;
|
|
that is, both <em>u</em> and <em>v</em> are reachable from each other. In other words, two
|
|
vertices of directed graph are in the same component if and only if they are
|
|
reachable from each other.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style10"><img alt="SSC example" src="Gifs/ssc4.gif" /> </p>
|
|
<p class="style10"><strong>
|
|
<span class="style15"> C<sub>1</sub>
|
|
C<sub>2</sub>
|
|
C<sub>3</sub>
|
|
C<sub>4</sub></span></strong></p>
|
|
<p class="style14"><span class="style1">The above directed graph has 4 strongly
|
|
connected components: C</span><sub><span class="style1">1</span></sub><span class="style1">,
|
|
C</span><sub><span class="style1">2</span></sub><span class="style1">, C</span><sub><span class="style1">3</span></sub><span class="style1">
|
|
and C</span><sub><span class="style1">4</span></sub><span class="style1">. If G
|
|
has an edge from some vertex in </span>C<sub><em><span class="style1">i</span></em></sub><span class="style1">
|
|
to some vertex in </span>C<sub><em><span class="style1">j</span></em></sub><span class="style1">
|
|
where </span><em><span class="style1">i</span></em><span class="style1"> ≠
|
|
</span><em><span class="style1">j</span></em><span class="style1">, then one can
|
|
reach any vertex in </span>C<sub><em><span class="style1">j</span></em></sub><span class="style1">
|
|
from any vertex in </span>C<sub><em><span class="style1">i</span></em></sub><span class="style1">
|
|
but not return. In the example, one can reach any vertex in C</span><sub><span class="style1">2</span></sub><span class="style1">
|
|
from any vertex in C</span><sub><span class="style1">1</span></sub><span class="style1">
|
|
but cannot return to C</span><sub><span class="style1">1</span></sub><span class="style1">
|
|
from C</span><sub><span class="style1">2</span></sub>.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1">The algorithm in CLRS for finding strongly connected
|
|
components of G = (V, E) uses the transpose of G, which define as:</p>
|
|
<ul class="style6">
|
|
<li>
|
|
<p class="style1">G<sup>T</sup> = (V, E<sup>T</sup>), where E<sup>T</sup> = {(<em>u</em>,
|
|
<em>v</em>): (<em>v</em>, <em>u</em>) in E}.</p>
|
|
</li>
|
|
<li>
|
|
<p class="style1">G<sup>T</sup> is G with all edges reversed.</p>
|
|
</li>
|
|
</ul>
|
|
<p class="style1">From the given graph G, one can create G<sup>T</sup> in linear time
|
|
(i.e., Θ(V + E)) if using adjacency lists.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Observation: </strong> </p>
|
|
<p class="style1">The graphs G and G<sup>T</sup> have the <span class="style4">same</span> SCC's. This means that
|
|
vertices <em>u</em> and <em>v</em> are reachable from each other in G if and only if reachable
|
|
from each other in G<sup>T</sup>.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Component Graph</strong></p>
|
|
<p class="style1">The idea behind the computation of SCC comes from a key
|
|
property of the component graph, which is defined as follows:</p>
|
|
<p class="style3">G<sup>SCC</sup> = (V<sup>SCC</sup>, E<sup>SCC</sup>), where V<sup>SCC</sup>
|
|
has one vertex for each SCC in G and E<sup>SCC </sup>has an edge if there's an
|
|
edge between the corresponding SCC's in G.</p>
|
|
<p class="style1">For our example (above) the G<sup>SCC</sup> is:</p>
|
|
<p class="style5"><img alt="SSC example" src="Gifs/ssc5.gif" /> </p>
|
|
<p class="style1"> </p>
|
|
<p class="style1">The key property of G<sup>SCC</sup> is that the component graph is a dag,
|
|
which the following lemma implies.</p>
|
|
<p class="style1"><strong>Lemma</strong> G<sup>SCC</sup> is a dag. More formally, let C and C' be
|
|
distinct SCC's in G, let u, v in C, u', v' in C', and suppose there is a path
|
|
<em>u</em>
|
|
<span class="style16">Þ</span> <em>u</em>' in G. Then there cannot also be a path v'
|
|
<span class="style16">Þ</span> v in G.</p>
|
|
<p class="style1"><strong>Proof</strong> Suppose there is a path
|
|
<em>v</em>' <span class="style16">Þ</span> <em>v</em> in G. Then there are paths <em>u</em>
|
|
<span class="style16">Þ</span> <em>u</em>' <span class="style16">Þ</span> <em>v</em>' and <em>v</em>'
|
|
<span class="style16">Þ</span> <em>v</em> <span class="style16">Þ</span> <em>u</em> in G. Therefore,
|
|
<em>u</em> and <em>v</em>' are reachable from each
|
|
other, so they are not in separate SCC's. </p>
|
|
<p class="style1">This completes the proof.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style7"><strong>ALGORITHM</strong></p>
|
|
<p class="style1">A DFS(G) produces a forest of DFS-trees. Let C be any strongly
|
|
connected component of G, let <em>v</em> be the first vertex on C discovered by
|
|
the DFS and let T be the DFS-tree containing <em>v</em> when DFS-visit(<em>v</em>)
|
|
is called all vertices in C are reachable from <em>v</em> along paths containing
|
|
visible vertices; DFS-visit(<em>v</em>) will visit every vertex in C, add it to
|
|
T as a descendant of <em>v</em>.</p>
|
|
<p class="style2">STRONGLY-CONNECTED-COMPONENTS (G)</p>
|
|
<p class="style2"> 1. <strong>Call</strong> DFS(G) to compute finishing
|
|
times f[u] for all <em>u</em>.<br />
|
|
2. <strong>Compute</strong> G<sup>T</sup><br />
|
|
3.<strong> Call</strong> DFS(G<sup>T</sup>), but in the main loop,
|
|
consider vertices in order of decreasing f[<em>u</em>] (as computed in first DFS)<br />
|
|
<strong> </strong>4.<strong> Output</strong> the vertices in each tree of
|
|
the depth-first forest formed in second DFS as a separate SCC.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Time</strong>: The algorithm takes linear time i.e.,
|
|
θ(V + E), to compute SCC of a digraph G.</p>
|
|
<p class="style1">From our <strong>Example</strong> (above): </p>
|
|
<p class="style11">1. Do DFS<br />
|
|
2. G<sup>T</sup><br />
|
|
3. DFS (roots blackened)</p>
|
|
<p class="style5"><img alt="SSC example" src="Gifs/ssc6.gif" /></p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Another Example</strong> (CLRS)
|
|
Consider a graph G = (V, E).</p>
|
|
<p class="style1">1. Call DFS(G)</p>
|
|
<p class="style5"><font size="4"><img border="0" src="Gifs/ssc1.gif" width="381" height="148"></font></p>
|
|
<p class="style1">2. Compute G<sup>T</sup></p>
|
|
<p class="style5"><font size="4"><img border="0" src="Gifs/ssc2.gif" width="379" height="149"/></font></p>
|
|
<p class="style1">3. Call DFS(G<sup>T</sup>) but this time consider the vertices
|
|
in order to decreasing finish time.</p>
|
|
<p class="style5">
|
|
<font size="4">
|
|
<img border="0" src="Gifs/ssc3.gif" width="322" height="108"/></font></p>
|
|
<p class="style1">4. Output the vertices of each tree in the DFS-forest as a
|
|
separate strongly connected components.</p>
|
|
<p class="style5">{<em>a</em>, <em>b</em>, <em>e</em>}, {<em>c</em>, <em>d</em>},
|
|
{<em>f</em>, <em>g</em>}, and {<em>h</em>}</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><span class="style8"><strong>Now the question is how can this possibly work</strong></span>?</p>
|
|
<p class="style1"><strong>Idea</strong> By considering vertices in second DFS in decreasing order of
|
|
finishing times from first DFS, we are visiting vertices of the component graph
|
|
in topological sort order.</p>
|
|
<p class="style1">To prove that it really works, first we deal with two notational
|
|
issues:</p>
|
|
<ul>
|
|
<li>
|
|
<p class="style1">We will be discussing d[<em>u</em>] and f[<em>u</em>]. These
|
|
always refer to the <span class="style4">first</span> DFS in the above algorithm.</p>
|
|
</li>
|
|
<li>
|
|
<p class="style1">We extend notation for <em>d</em> and <em>f</em> to sets of
|
|
vertices U subset V:</p>
|
|
<ul>
|
|
<li>
|
|
<p class="style1">d(U) = min<sub>u in U</sub>
|
|
{d[<em>u</em>]}
|
|
(earliest discovery time of any vertex in U)</p>
|
|
</li>
|
|
<li>
|
|
<p class="style1">f(U) = min<sub>u in U</sub>
|
|
{f[<em>u</em>]}
|
|
(latest finishing time of any vertex in U)</p>
|
|
</li>
|
|
</ul>
|
|
</li>
|
|
</ul>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Lemma</strong> Let C and C' be distinct SCC's in G = (V, E). Suppose there is
|
|
an edge (<em>u</em>, <em>v</em>) in E such that <em>u</em> in C and
|
|
<em>v</em> in C'.
|
|
Then f(C) > f(C').</p>
|
|
<p class="style5"><img alt="scc6-Lemma1" src="Gifs/scc6-lemma.gif" /> </p>
|
|
<p class="style1"><strong>Proof</strong> There are two cases, depending on which SCC had the first
|
|
discovered vertex during the first DFS.</p>
|
|
<p class="style1"><strong>Case i.</strong> If d(C) > d(C'), let <em>x</em> be the first vertex discovered
|
|
in C. At time d[<em>x</em>], all vertices in C and C' are white. Thus, there exist paths
|
|
of white vertices from <em>x</em> to all vertices in C and C'.</p>
|
|
<p class="style1">By the white-path theorem, all vertices in C and C' are
|
|
descendants of <em>x</em> in depth-first tree.</p>
|
|
<p class="style1">By the parenthesis theorem, we have f[<em>x</em>] = f(C) > f(C').</p>
|
|
<p class="style1"><strong>Case ii.</strong> If d(C) > d(C'), let <em>y</em> be the first vertex discovered
|
|
in C'. At time d[<em>y</em>], all vertices in C' are white and there is a white path from
|
|
<em>y</em> to each vertex in C. This implies that all vertices in C' become descendants
|
|
of <em>y</em>. Again, f[y] = f(C').</p>
|
|
<p class="style1">At time d[<em>y</em>], all vertices in C are white.</p>
|
|
<p class="style1">By earlier lemma, since there is an edge (<em>u</em>, <em>v</em>), we cannot
|
|
have a path from C' to C. So, no vertex in C is reachable from <em>y</em>. Therefore, at
|
|
time f[<em>y</em>], all vertices in C are still white. Therefore, for all <em>w</em> in C, f[<em>w</em>] >
|
|
f[<em>y</em>], which implies that f(C) > f(C').</p>
|
|
<p class="style1">This completes the proof.</p>
|
|
<p class="style1"><strong>Corollary</strong> Let C and C' be distinct SCC's in G = (V, E). Suppose there is
|
|
an edge (<em>u</em>, <em>v</em>) in E<sup>T</sup> where <em>u</em> in C and
|
|
<em>v</em> in C'. Then f(C) <
|
|
f(C').</p>
|
|
<p class="style1"><strong>Proof</strong> Edge (<em>u</em>, <em>v</em>)
|
|
in E<sup>T</sup> implies (<em>v</em>,
|
|
<em>u</em>) in E.
|
|
Since SCC's of G and G<sup>T</sup> are the same, f(C') > f(C). This completes
|
|
the proof.</p>
|
|
<p class="style1"><strong>Corollary</strong> Let C and C' be distinct SCC's in G = (V, E), and suppose that
|
|
f(C) > f(C'). Then there cannot be an edge from C to C' in G<sup>T</sup>.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Proof Idea</strong> It's the contrapositive of the previous corollary.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style9"><strong>Now, we have the intuition to understand why the SCC procedure
|
|
works.</strong></p>
|
|
<p class="style1">When we do the second DFS, on G<sup>T</sup>, start with SCC C
|
|
such that f(C) is maximum. The second DFS starts from some <em>x</em> in C, and it visits
|
|
all vertices in C. Corollary says that since f(C) > f(C') for all C'
|
|
<span class="style17">≠</span> C, there
|
|
are no edges from C to C' in G<sup>T</sup>. Therefore, DFS will visit only vertices in C.</p>
|
|
<p class="style1">Which means that the depth-first tree rooted at <em>x</em> contains
|
|
<span class="style4">exactly</span> the vertices of C.</p>
|
|
<p class="style1">The next root chosen in the second DFS is in SCC C' such that
|
|
f(C') is maximum over all SCC's other than C. DFS visits all vertices in C', but
|
|
the only edges out of C' go to C, <span class="style4">which we've already visited</span>.</p>
|
|
<p class="style1">Therefore, the only tree edges will be to vertices in C'.</p>
|
|
<p class="style1">We can continue the process.</p>
|
|
<p class="style1">Each time we choose a root for the second DFS, it can reach
|
|
only</p>
|
|
<ul>
|
|
<li>
|
|
<p class="style1">vertices in its SCC ‾ get tree edges to these,</p>
|
|
</li>
|
|
<li>
|
|
<p class="style1">vertices in SCC's <span class="style4">already visited</span> in second
|
|
DFS ‾ get <span class="style4">no</span> tree edges to these.</p>
|
|
</li>
|
|
</ul>
|
|
<p class="style1">We are visiting vertices of (G<sup>T</sup>)<sup>SCC</sup> in reverse of
|
|
topologically sorted order. [CLRS has a formal proof.]</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1">Before leaving strongly connected components, lets prove that
|
|
the component graph of G = (V, E) is a directed acyclic graph.</p>
|
|
<p class="style1"><br />
|
|
<strong>Proof</strong> (by contradiction) Suppose component
|
|
graph of G = (V, E) was not a DAG and G comprised of a cycle consisting of
|
|
vertices v<sub>1</sub>, v<sub>2</sub> , . . . , v<sub>n</sub> . Each <em>v<sub>i</sub></em>
|
|
corresponds to a strongly connected component (SCC) of component graph G. If v<sub>1</sub>,
|
|
v<sub>2</sub> , . . . , v<sub>n</sub> themselves form a cycle then each <em>v<sub>i</sub></em>
|
|
( <em>i</em> runs from 1 to <em>n</em>) should have been included in the SCC
|
|
corresponding to <em>v<sub>j</sub></em> ( <em>j</em> runs from 1 to <em>n</em>
|
|
and <em>i <span class="style17">≠</span> j</em>). But each of the vertices is a vertex from a
|
|
difference SCC of G. Hence, we have a contradiction! Therefore, SCC of G is a
|
|
directed acyclic graph.<br />
|
|
</p>
|
|
<p class="style1"> </p>
|
|
<p class="style1"><strong>Related Problems</strong></p>
|
|
<p class="style11">1. Edge-vertex connectivity problem.<br />
|
|
2. Shortest path problem.</p>
|
|
<p class="style1"> </p>
|
|
<p class="style5"><font size="4"><img SRC="../../../Maingif/redline.gif" height=2 width=640/></font></p>
|
|
|
|
<p class="style5">
|
|
|
|
<a href="../../algorithm.html">
|
|
|
|
<font size="4">
|
|
|
|
<img src="../../../Maingif/back.gif" border=0 height=47 width=49/></font></a></p>
|
|
<p class="style12">Updated: March 13, 2010.</p>
|
|
|
|
</body>
|
|
|
|
</html>
|