13.07.2015 Views

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Complexity of Connected Components <strong>in</strong> Evolv<strong>in</strong>g Graphs 269the arc is present both at times f a (u j ,v j ) − 1 and f a (u j ,v j ), s<strong>in</strong>ce w(u j ,v j ) < 1,the packet will reach v j <strong>in</strong> t uj + w(u j ,v j ). If, however, the arc is not present attime f a (u j ,v j ) − 1, then the transmission process itself starts at the f a (u j ,v j ) thstep (i.e. from time f a (u j ,v j ) − 1 to time f a (u j ,v j )), thus reach<strong>in</strong>g v j <strong>by</strong> timef a (u j ,v j ) − 1+w(u j ,v j ).We remark that a rooted directed tree can also be computed over an o-SCC V G ′. As a modification for that purpose, V G must be replaced <strong>by</strong> V G ′ andcorrespond<strong>in</strong>gly, Step 3 of Algorithm 1 should be modified to V G ′ ⊂ V f s<strong>in</strong>ce thefragment can also conta<strong>in</strong> the h-nodes for the vertices <strong>in</strong> V G ′ and the loop canstop once all the vertices are covered.Algorithm 1 is a greedy algorithm that always chooses the arc that transmits<strong>in</strong> m<strong>in</strong>imum time. The proof of its correctness is the same as the proof of thePrim-Dijkstra algorithm [4]. If the maximum outdegree of each vertex is D, theneach step of <strong>in</strong>creas<strong>in</strong>g the fragment will take O(NDlog T ) time and the fragmentwill <strong>in</strong>crease N times add<strong>in</strong>g up to a total execution time of O(N 2 D log T )steps.5 ConclusionThe two important results <strong>in</strong> this paper are the <strong>in</strong>tractability of the decomposition<strong>in</strong>to (open) strongly connected components <strong>in</strong> FSDN’s and the constructionof DMST’s over an already exist<strong>in</strong>g strongly connected components.The first result implies that it is possible to lead a non-strongly connectednetwork towards strong connectedness <strong>by</strong> add<strong>in</strong>g <strong>in</strong>termediary agents to serveas hops between two nodes that are out of range from each other. An <strong>in</strong>terest<strong>in</strong>gproblem would be to f<strong>in</strong>d a way to add such l<strong>in</strong>ks so as to m<strong>in</strong>imize the numberof <strong>in</strong>termediary (help<strong>in</strong>g) nodes. Another way for further research is to designapproximation algorithms for (open) strongly connected components <strong>in</strong> evolv<strong>in</strong>gdigraphs.AcknowledgmentsThe authors are grateful to Aub<strong>in</strong> Jarry and Stephane Perennes for very fruitfuldiscussions.References1. A. Borod<strong>in</strong> and R. El-Yaniv. Onl<strong>in</strong>e computation and competitive analysis. CambridgeUniversity Press, 1998.2. B. Bui-Xuan, A. Ferreira, and A. Jarry. Comput<strong>in</strong>g shortest, fastest, and foremostjourneys <strong>in</strong> dynamic networks. International Journal of Foundations of <strong>Computer</strong><strong>Science</strong>, 14(2):267–285, April 2003.3. Y. J. Chu and T. H. Liu. On the shortest arborescence of a directed graph. <strong>Science</strong>S<strong>in</strong>ica, 14:1396–1400, 1965.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!