DocumentCode :
2477785
Title :
Bit allocation in sub-linear time and the multiple-choice knapsack problem
Author :
Mohr, Alexander E.
Author_Institution :
Dept. of Comput. Sci. & Eng., Washington Univ., Seattle, WA, USA
fYear :
2002
fDate :
2002
Firstpage :
352
Lastpage :
361
Abstract :
We show that the problem of optimal bit allocation among a set of independent discrete quantizers given a budget constraint is equivalent to the multiple choice knapsack problem (MCKP). This result has three implications: first, it provides a trivial proof that the problem of optimal bit allocation is NP-hard and that its related decision problem is NP-complete; second, it unifies research into solving these problems that has to date been done independently in the data compression community and the operations research community; third, many practical algorithms for approximating the optimal solution to MCKP can be used for bit allocation. We implement the GBFOS, partition-search, and Dudzinski-Walukiewicz algorithms and compare their running times for a variety of problem sizes.
Keywords :
computational complexity; constraint theory; data compression; knapsack problems; optimisation; quantisation (signal); search problems; Dudzinski-Walukiewicz algorithm; GBFOS algorithm; NP-complete problem; NP-hard problem; approximation; budget constraint; data compression; decision problem; independent discrete quantizers; multiple choice knapsack problem; optimal bit allocation; partition-search algorithm; running times; sub-linear time; Approximation algorithms; Bit rate; Character generation; Computer science; Data compression; Equations; Operations research; Partitioning algorithms;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Data Compression Conference, 2002. Proceedings. DCC 2002
ISSN :
1068-0314
Print_ISBN :
0-7695-1477-4
Type :
conf
DOI :
10.1109/DCC.2002.999973
Filename :
999973
Link To Document :
بازگشت