research article on a property of a three-dimensional...
TRANSCRIPT
Hindawi Publishing CorporationJournal of Discrete MathematicsVolume 2013 Article ID 797249 3 pageshttpdxdoiorg1011552013797249
Research ArticleOn a Property of a Three-Dimensional Matrix
David Blokh
C D Technologies Ltd Israel
Correspondence should be addressed to David Blokh david blokh012netil
Received 11 June 2013 Accepted 24 September 2013
Academic Editor Hong J Lai
Copyright copy 2013 David BlokhThis is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Let 119878119899be the symmetrical group acting on the set 1 2 119899 and 119909 119910 isin 119878
119899 Consider the set 119882 = (119894 119909(119894) 119910(119894)) | 1 le 119894 le 119899
|119894 minus 119909(119894)| gt 1 or |119894 minus 119910(119894)| gt 1 or |119909(119894) minus 119910(119894)| gt 1The main result of this paper is the following theorem If the number of119882 setentries is more than [1198993] then there exist entries (119894
1 119909(1198941) 119910(1198941)) (1198942 119909(1198942) 119910(1198942)) (1198943 119909(1198943) 119910(1198943)) isin 119882 such that |119894
1minus 119909(1198942)| le 1
|1198941minus119910(1198943)| le 1 and |119909(119894
2) minus 119910(119894
3)| le 1 The application of this theorem to the three-dimensional assignment problem is considered
1 Introduction
Let 1198773119899be the set of 119899 times 119899 times 119899-matrices over the field of real
numbersThree-dimensional matrix not only is an interesting
mathematical object [1ndash3] but also has applications in manyfields such as theoretical physics [4] and operational research[5 6]
Let 119878119899be the symmetrical group acting on the set
1 2 119899 119909 119910 isin 119878119899 119886119894119895119896 isin 1198773
119899and
119866 = 119886119894119895119896|1003816100381610038161003816119894 minus 119895
1003816100381610038161003816 le 1 and |119894 minus 119896| le 1 and1003816100381610038161003816119895 minus 119896
1003816100381610038161003816 le 1
119867 = 119886119894119895119896|1003816100381610038161003816119894 minus 119895
1003816100381610038161003816 gt 1 or |119894 minus 119896| gt 1 or1003816100381610038161003816119895 minus 119896
1003816100381610038161003816 gt 1
119879 = 119886119894119909(119894)119910(119894)
| 1 le 119894 le 119899
(1)
The main result of this paper is the following theorem
Theorem 1 If 119886119894119895119896 isin 119877
3
119899 119909 119910 isin 119878
119899and the number of 119879 cap
119867 set entries is more than [1198993] then there exist entries 119886119894111989511198961
119886119894211989521198962
and 119886119894311989531198963
isin 119879 cap 119867 such that the entry 119886119894111989521198963
isin 119866
We give another formulation of Theorem 1Consider the set119882 = (119894 119909 (119894) 119910 (119894)) | 1 le 119894 le 119899 |119894 minus 119909 (119894)|
gt 1 or1003816100381610038161003816119894 minus 119910 (119894)
1003816100381610038161003816 gt 1 or1003816100381610038161003816119909 (119894) minus 119910 (119894)
1003816100381610038161003816 gt 1
(2)
Theorem 2 If the number of 119882 set entries is more than[1198993] then there exist entries (119894
1 119909(1198941) 119910(1198941)) (1198942 119909(1198942) 119910(1198942))
(1198943 119909(1198943) 119910(1198943)) isin 119882 such that |119894
1minus 119909(1198942)| le 1 |119894
1minus 119910(1198943)| le 1
and |119909(1198942) minus 119910(119894
3)| le 1
2 Proof of Theorem 1
We prove the theorem by contradiction The set of matrix119886119894119895119896 isin 119877
3
119899entries with one index fixed and the two others
having values from 1 to 119899 will be called a layer We denote alayer by 119871119903
119904 where 119903 indicates the location of a fixed index and
119904 indicates its value For example 1198712119904= 119886119894119904119896| 119894 119896 = 1 119899
Furthermore entries from119866will be called basic entries from119867 will be called nonbasic 119879 will be termed a trajectory thelayer containing a basic trajectory entry will be termed a basiclayer and the layer containing a nonbasic trajectory entrywillbe termed a nonbasic layer
If 1198711119904is a nonbasic layer then one layer in the pair
1198712
119904minus1 1198713119904minus1
is basicSuppose to the contrary that layers 1198711
119904 1198712119904minus1
and 1198713119904minus1
arenonbasic Let 119886
11990411989511198961
1198861198942119904minus1119896
2
and 11988611989431198953119904minus1
be the trajectoryentries of 1198711
119904 1198712119904minus1
1198713119904minus1
layers Replace 11988611990411989511198961
1198861198942119904minus1119896
2
and11988611989431198953119904minus1
with 119886119904119904minus1119904minus1
119886119894211989511198961
and 119886119894311989531198962
The nonbasic tra-jectory entries 119886
11990411989511198961
1198861198942119904minus1119896
2
and 11988611989431198953119904minus1
are replaced withentries 119886
119904119904minus1119904minus1 119886119894211989511198961
and 119886119894311989531198962
among which there is abasic one
Similar assertions may be proved for the 1198711119904layer and
the following layer pairs 1198712119904minus1 1198713
119904 1198712119904 1198713
119904minus1 1198712119904 1198713
119904 1198712119904 1198713
119904+1
1198712
119904+1 1198713
119904 and 1198712
119904+1 1198713
119904+1
Three nonbasic layers may not be consecutive
2 Journal of Discrete Mathematics
Suppose that layers 1198711119904minus1
1198711119904 and 1198711
119904+1are nonbasic One
of the layers 1198712119904or 1198713119904is basic In these layers the basic
entries may be 119886119904minus1119904119904minus1
119886119904minus1119904119904
119886119904119904119904minus1
119886119904119904119904 119886119904119904119904+1
119886119904+1119904119904
and 119886
119904+1119904119904+1for the layer 1198712
119904 and 119886
119904minus1119904minus1119904 119886119904minus1119904119904
119886119904119904minus1119904
119886119904119904119904 119886119904+1119904119904
119886119904119904+1119904
119886119904+1119904+1119904
for the layer 1198713119904 However this
contradicts the assumption that the layers 1198711119904minus1
1198711119904 and 1198711
119904+1
are nonbasicThe fact that the two first and the two last layersmay not be nonbasic is proved in a similar way
All assertions given below represent conditions thatprevent replacing nonbasic trajectory entries with entries thatinclude a basic one
Consider the sequence of layers 11987111 1198711
2 119871
1
119904minus1 1198711
119904
1198711
119904+1 119871
1
119899minus1 and 1198711
119899 Given below are possible arrange-
ments of layers in this sequence A basic layer 1198711119904is denoted by
1 a nonbasic layer 1198711119904is denoted by 0
Consider
(1) 11100111 1100111 1110011 110011
(2) 111010111 01111010 01011110 001111010 010111100 0101111010
(3) 0111100 00111100 0011110
(4) 0111 1110 10111 11101 011110
We prove the first arrangement of item 1 Suppose thatthe two nonbasic layers are followed only by two basic layersthat is 00110 Let there be layers 1198711
119904 1198711119904+1
1198711119904+2
1198711119904+3
1198711
119904+4 Consider layer pairs1198712
119904+11198713119904+1
1198712119904+1
1198713119904+2
1198712119904+2
1198713119904+1
1198712119904+2
and 1198713
119904+2 At least one layer of each pair is basic These four
pairs of layers contain at least two basic entries and theseentries are located in layers 1198711
119904+2and 1198711
119904+3 The pair 1198712
119904+3 1198713119904+3
includes a basic layer The first coordinate of the basic layerentry may be 119904 + 2 119904 + 3 or 119904 + 4 But the 1198711
119904+2and 1198711
119904+3
layers are occupied while the 1198711119904+4
layer is nonbasic Thereis a contradiction
The other arrangements are proved in a similar wayThus nonbasic layers may not be arranged closer than
those in the above variants But these variants do not allow forthe composition of a combination containingmore than [1198993]nonbasic layers Hence it follows that if a trajectory includesmore than [1198993] nonbasic entries then one of the variants isviolated and nonbasic trajectory entries can be replaced by aset of entries containing a basic entry
3 Application to the Three-DimensionalAssignment Problem
The three-dimensional assignment problem (AP3) is animportant combinatorial optimization problem It is suf-ficient to note that the particular case of AP3 the 3-dimensional matching problem is one of the six main NP-hard problems [7] The formal AP3 statement is as followsfor a matrix 119886
119894119895119896 isin 119877
3
119899 find permutations 119909 119910 isin 119878
119899such
that sum119899119894=1119886119894119909(119894)119910(119894)
is maximized
In this paper the AP3 is considered for a special class of119899 times 119899 times 119899-matrices 119880(119899) A matrix
10038171003817100381710038171003817119886119894119895119896
10038171003817100381710038171003817isin 119880 (119899) if 119886
119894119895119896ge 0
min119886119894119895119896isin119866
119886119894119895119896ge 119862 sdot max119886119894119895119896isin119867
119886119894119895119896 119862 ge 3
(3)
One of AP3 interpretations is the following There are 119899employees and two job sets of 119899 jobs each If the 119894th employeeperforms the 119895th job of the first set and the 119896th job of the sec-ond set then the effect equals 119886
119894119895119896 It is required to distribute
the jobs among the employees in such away so as tomaximizethe total effect
Let us describe the situation that will lead to the AP3 formatrices from 119880(119899) As a rule the employees are orderedby qualification while the jobs are ordered by complexityA higher effect is reached when a more qualified employeeperforms a more complex job and we may arrive at the AP3for matrices from 119880(119899)
The particular cases of AP3 [5 8ndash11] the nonpolynomialexact algorithms for the AP3 [6 12] and heuristics for theAP3 [13 14] were considered
The NP-hard particular cases of the traveling salesmanproblem with sets of matrices whose structures are similar tothose ofmatrices from119880(119899) have been considered previously[15 16]
Theorem 3 The AP3 for matrices from 119880(119899) is NP-hard
Proof Let 119887119894119895119896 isin 1198773
119899 119887119894119895119896ge 0 and numbers119860119861 are such that
119860 = 3 sdotmax 119887119894119895119896 119861 = 119899119860
Consider 4119899 times 4119899 times 4119899-matrix 119886119894119895119896 with nonnegative
entries
1198864119904minus34119904minus24119904minus2
= 119861 119904 = 1 119899
1198864119904minus24119904minus14119904minus1
= 119861 119904 = 1 119899
119886411990441199044119904
= 119861 119904 = 1 119899
119886119903119906V = 119860 119903 119906 V = 1 119899 if |119903 minus 119906| le 1
|119903 minus V| le 1 |119906 minus V| le 1 119886119903119906V = 119861
1198864119894minus14119895minus34119896minus3
= 119887119894119895119896
119894 119895 119896 = 1 119899
119886119894119895119896= 0 in other cases
(4)
The matrix 119886119894119895119896 isin 119880(4119899)
All entries that are equal to 119861 belong to the optimal solu-tion of AP3 for 119886
119894119895119896matrix This signifies that if the optimal
solution of AP3 for 119886119894119895119896 matrix is known then the optimal
solution of AP3 for 119887119894119895119896matrix is known as well Therefore
the AP3 for arbitrary matrices is polynomially reducible tothe AP3 for matrices from 119880(119899)
Corollary 4 If 119886119894119895119896 isin 119880(119899) then a set T exists on which the
AP3 optimum is reached and the number of 119879 cap 119867 set entriesis no more than [1198993]
Indeed if for optimal T of matrix 119886119894119895119896 isin 119880(119899) the number
of 119879 cap 119867 set entries is more than [1198993] then according to
Journal of Discrete Mathematics 3
Theorem 1 it is possible to make such replacement of entriessuch that the sum of T set entries will not change and thenumber of 119879 cap 119867 set entries is reduced
Theorem 5 If 119886119894119895119896 isin 119880(119899) then the AP3 optimum for a
1198861015840
119894119895119896 matrix where 1198861015840
119894119895119896= 119886119894119895119896
for 1198861015840119894119895119896isin 119866 and 1198861015840
119894119895119896= 0 for
1198861015840
119894119895119896isin 119867 represents an approximation for the AP3 optimum of
the 119886119894119895119896matrix with a relative error not exceeding 1(3C)
Proof Let 119886119894119895119896 isin 119880(119899) 119898
1is the AP3 optimum for 119886
119894119895119896
119891 119892 isin 119878119899are such that119898
1= sum119899
119894=1119886119894119891(119894)119892(119894)
and the number of119886119894119891(119894)119892(119894)
| 119894 = 1 119899 cap 119867 set entries is no more than [1198993]1198982is the sum of 119886
119894119891(119894)119892(119894)| 119894 = 1 119899 cap 119867 set entries and
1198983= 1198981minus 11989821198981015840 is the AP3 optimum for 1198861015840
119894119895119896
Note that1198983= sum119899
119894=11198861015840
119894119891(119894)119892(119894)and1198981015840 ge 119898
3
Insofar as 119886119894119895119896ge 1198861015840
119894119895119896 119894 119895 119896 = 1 119899 then 119898
1ge 1198981015840
Hence1198983le 1198981015840
le 1198981and (119898
1minus 1198981015840
)1198981le (1198981minus 1198983)1198981=
11989821198981
Since1198982le (13)119899sdotmax
119886119894119895119896isin119867119886119894119895119896
and1198981ge 119899sdotmin
119886119894119895119896isin119866119886119894119895119896
then 11989821198981le 1(3119862) Hence (119898
1minus 1198981015840
)1198981le 1(3119862) and
the relative error of1198981015840 taken as the AP3 approximate solutionfor the 119886
119894119895119896matrix is no more than 1(3119862)
Remark 6 An exact119874(119899)-algorithmofAP3 solution has beenconstructed [9] for 119899 times 119899 times 119899-matrices 119886
119894119895119896 such that 119886
119894119895119896ge
0 for 119886119894119895119896isin 119866 and 119886
119894119895119896= 0 for 119886
119894119895119896isin 119867 using dynamical
programming
Acknowledgments
The author is grateful to Professor Y Dinitz and Professor GGutin for their attention to this work and valuable commentsAlso the author is grateful to the anonymous referee of thisjournal for the valuable comments and suggestions
References
[1] R A Brualdi and J Csima ldquoOn the plane term rank of a threedimensional matrixrdquo Proceedings of the AmericanMathematicalSociety vol 54 no 1 pp 471ndash473 1976
[2] J Csima ldquoOn the plane term rank of three dimensional matri-cesrdquo Discrete Mathematics vol 28 no 2 pp 147ndash152 1979
[3] D K Kim Y A Kim and K Park ldquoGeneralizations of suffixarrays to multi-dimensional matricesrdquo Theoretical ComputerScience vol 302 no 1ndash3 pp 401ndash416 2003
[4] Y Kawamura ldquoCubic matrices generalized spin algebra anduncertainty relationrdquo Progress ofTheoretical Physics vol 110 no3 pp 579ndash587 2003
[5] R E Burkard R Rudolf and G J Woeginger ldquoThree-dimen-sional axial assignment problems with decomposable cost coef-ficientsrdquoDiscrete Applied Mathematics vol 65 no 1ndash3 pp 123ndash139 1996
[6] W P Pierskalla ldquoThe multidimensional assignment problemrdquoOperations Research vol 16 no 2 pp 422ndash431 1968
[7] M R Garey and D C Johnson Computer and Intractability AGuide To the Theory of NP-Completeness W H Freeman SanFrancisco Calif USA 1979
[8] W W Bein P Brucker J K Park and P K Pathak ldquoA Mongeproperty for the d-dimensional transportation problemrdquo Dis-crete Applied Mathematics vol 58 no 2 pp 97ndash109 1995
[9] D A Blokh ldquoAssignment problem for three-index Jacobimatri-cesrdquoAutomation and Remote Control vol 50 no 5 pp 672ndash6771989
[10] Y Crama and F C R Spieksma ldquoApproximation algorithms forthree-dimensional assignment problems with triangle inequal-itiesrdquo European Journal of Operational Research vol 60 no 3pp 273ndash279 1992
[11] F C R Spieksma and G J Woeginger ldquoGeometric three-dimensional assignment problemsrdquo European Journal of Oper-ational Research vol 91 no 3 pp 611ndash618 1996
[12] E Balas and M J Saltzman ldquoAn algorithm for the three-indexassignment problemrdquo Operations Research vol 39 no 1 pp150ndash161 1991
[13] G Gutin B Goldengorin and J Huang ldquoWorst case analysisof Max-Regret greedy and other heuristics for multidimen-sional assignment and traveling salesman problemsrdquo Journal ofHeuristics vol 14 no 2 pp 169ndash181 2008
[14] B-J Kim W L Hightower P M Hahn Y-R Zhu and L SunldquoLower bounds for the axial three-index assignment problemrdquoEuropean Journal of Operational Research vol 202 no 3 pp654ndash668 2010
[15] D Blokh and G Gutin ldquoMaximizing traveling salesman prob-lem for special matricesrdquo Discrete Applied Mathematics vol 56no 1 pp 83ndash86 1995
[16] D Blokh and E Levner ldquoAn approximation algorithm withperformance guarantees for the maximum traveling salesmanproblemon specialmatricesrdquoDiscrete AppliedMathematics vol119 no 1-2 pp 139ndash148 2002
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Discrete Mathematics
Suppose that layers 1198711119904minus1
1198711119904 and 1198711
119904+1are nonbasic One
of the layers 1198712119904or 1198713119904is basic In these layers the basic
entries may be 119886119904minus1119904119904minus1
119886119904minus1119904119904
119886119904119904119904minus1
119886119904119904119904 119886119904119904119904+1
119886119904+1119904119904
and 119886
119904+1119904119904+1for the layer 1198712
119904 and 119886
119904minus1119904minus1119904 119886119904minus1119904119904
119886119904119904minus1119904
119886119904119904119904 119886119904+1119904119904
119886119904119904+1119904
119886119904+1119904+1119904
for the layer 1198713119904 However this
contradicts the assumption that the layers 1198711119904minus1
1198711119904 and 1198711
119904+1
are nonbasicThe fact that the two first and the two last layersmay not be nonbasic is proved in a similar way
All assertions given below represent conditions thatprevent replacing nonbasic trajectory entries with entries thatinclude a basic one
Consider the sequence of layers 11987111 1198711
2 119871
1
119904minus1 1198711
119904
1198711
119904+1 119871
1
119899minus1 and 1198711
119899 Given below are possible arrange-
ments of layers in this sequence A basic layer 1198711119904is denoted by
1 a nonbasic layer 1198711119904is denoted by 0
Consider
(1) 11100111 1100111 1110011 110011
(2) 111010111 01111010 01011110 001111010 010111100 0101111010
(3) 0111100 00111100 0011110
(4) 0111 1110 10111 11101 011110
We prove the first arrangement of item 1 Suppose thatthe two nonbasic layers are followed only by two basic layersthat is 00110 Let there be layers 1198711
119904 1198711119904+1
1198711119904+2
1198711119904+3
1198711
119904+4 Consider layer pairs1198712
119904+11198713119904+1
1198712119904+1
1198713119904+2
1198712119904+2
1198713119904+1
1198712119904+2
and 1198713
119904+2 At least one layer of each pair is basic These four
pairs of layers contain at least two basic entries and theseentries are located in layers 1198711
119904+2and 1198711
119904+3 The pair 1198712
119904+3 1198713119904+3
includes a basic layer The first coordinate of the basic layerentry may be 119904 + 2 119904 + 3 or 119904 + 4 But the 1198711
119904+2and 1198711
119904+3
layers are occupied while the 1198711119904+4
layer is nonbasic Thereis a contradiction
The other arrangements are proved in a similar wayThus nonbasic layers may not be arranged closer than
those in the above variants But these variants do not allow forthe composition of a combination containingmore than [1198993]nonbasic layers Hence it follows that if a trajectory includesmore than [1198993] nonbasic entries then one of the variants isviolated and nonbasic trajectory entries can be replaced by aset of entries containing a basic entry
3 Application to the Three-DimensionalAssignment Problem
The three-dimensional assignment problem (AP3) is animportant combinatorial optimization problem It is suf-ficient to note that the particular case of AP3 the 3-dimensional matching problem is one of the six main NP-hard problems [7] The formal AP3 statement is as followsfor a matrix 119886
119894119895119896 isin 119877
3
119899 find permutations 119909 119910 isin 119878
119899such
that sum119899119894=1119886119894119909(119894)119910(119894)
is maximized
In this paper the AP3 is considered for a special class of119899 times 119899 times 119899-matrices 119880(119899) A matrix
10038171003817100381710038171003817119886119894119895119896
10038171003817100381710038171003817isin 119880 (119899) if 119886
119894119895119896ge 0
min119886119894119895119896isin119866
119886119894119895119896ge 119862 sdot max119886119894119895119896isin119867
119886119894119895119896 119862 ge 3
(3)
One of AP3 interpretations is the following There are 119899employees and two job sets of 119899 jobs each If the 119894th employeeperforms the 119895th job of the first set and the 119896th job of the sec-ond set then the effect equals 119886
119894119895119896 It is required to distribute
the jobs among the employees in such away so as tomaximizethe total effect
Let us describe the situation that will lead to the AP3 formatrices from 119880(119899) As a rule the employees are orderedby qualification while the jobs are ordered by complexityA higher effect is reached when a more qualified employeeperforms a more complex job and we may arrive at the AP3for matrices from 119880(119899)
The particular cases of AP3 [5 8ndash11] the nonpolynomialexact algorithms for the AP3 [6 12] and heuristics for theAP3 [13 14] were considered
The NP-hard particular cases of the traveling salesmanproblem with sets of matrices whose structures are similar tothose ofmatrices from119880(119899) have been considered previously[15 16]
Theorem 3 The AP3 for matrices from 119880(119899) is NP-hard
Proof Let 119887119894119895119896 isin 1198773
119899 119887119894119895119896ge 0 and numbers119860119861 are such that
119860 = 3 sdotmax 119887119894119895119896 119861 = 119899119860
Consider 4119899 times 4119899 times 4119899-matrix 119886119894119895119896 with nonnegative
entries
1198864119904minus34119904minus24119904minus2
= 119861 119904 = 1 119899
1198864119904minus24119904minus14119904minus1
= 119861 119904 = 1 119899
119886411990441199044119904
= 119861 119904 = 1 119899
119886119903119906V = 119860 119903 119906 V = 1 119899 if |119903 minus 119906| le 1
|119903 minus V| le 1 |119906 minus V| le 1 119886119903119906V = 119861
1198864119894minus14119895minus34119896minus3
= 119887119894119895119896
119894 119895 119896 = 1 119899
119886119894119895119896= 0 in other cases
(4)
The matrix 119886119894119895119896 isin 119880(4119899)
All entries that are equal to 119861 belong to the optimal solu-tion of AP3 for 119886
119894119895119896matrix This signifies that if the optimal
solution of AP3 for 119886119894119895119896 matrix is known then the optimal
solution of AP3 for 119887119894119895119896matrix is known as well Therefore
the AP3 for arbitrary matrices is polynomially reducible tothe AP3 for matrices from 119880(119899)
Corollary 4 If 119886119894119895119896 isin 119880(119899) then a set T exists on which the
AP3 optimum is reached and the number of 119879 cap 119867 set entriesis no more than [1198993]
Indeed if for optimal T of matrix 119886119894119895119896 isin 119880(119899) the number
of 119879 cap 119867 set entries is more than [1198993] then according to
Journal of Discrete Mathematics 3
Theorem 1 it is possible to make such replacement of entriessuch that the sum of T set entries will not change and thenumber of 119879 cap 119867 set entries is reduced
Theorem 5 If 119886119894119895119896 isin 119880(119899) then the AP3 optimum for a
1198861015840
119894119895119896 matrix where 1198861015840
119894119895119896= 119886119894119895119896
for 1198861015840119894119895119896isin 119866 and 1198861015840
119894119895119896= 0 for
1198861015840
119894119895119896isin 119867 represents an approximation for the AP3 optimum of
the 119886119894119895119896matrix with a relative error not exceeding 1(3C)
Proof Let 119886119894119895119896 isin 119880(119899) 119898
1is the AP3 optimum for 119886
119894119895119896
119891 119892 isin 119878119899are such that119898
1= sum119899
119894=1119886119894119891(119894)119892(119894)
and the number of119886119894119891(119894)119892(119894)
| 119894 = 1 119899 cap 119867 set entries is no more than [1198993]1198982is the sum of 119886
119894119891(119894)119892(119894)| 119894 = 1 119899 cap 119867 set entries and
1198983= 1198981minus 11989821198981015840 is the AP3 optimum for 1198861015840
119894119895119896
Note that1198983= sum119899
119894=11198861015840
119894119891(119894)119892(119894)and1198981015840 ge 119898
3
Insofar as 119886119894119895119896ge 1198861015840
119894119895119896 119894 119895 119896 = 1 119899 then 119898
1ge 1198981015840
Hence1198983le 1198981015840
le 1198981and (119898
1minus 1198981015840
)1198981le (1198981minus 1198983)1198981=
11989821198981
Since1198982le (13)119899sdotmax
119886119894119895119896isin119867119886119894119895119896
and1198981ge 119899sdotmin
119886119894119895119896isin119866119886119894119895119896
then 11989821198981le 1(3119862) Hence (119898
1minus 1198981015840
)1198981le 1(3119862) and
the relative error of1198981015840 taken as the AP3 approximate solutionfor the 119886
119894119895119896matrix is no more than 1(3119862)
Remark 6 An exact119874(119899)-algorithmofAP3 solution has beenconstructed [9] for 119899 times 119899 times 119899-matrices 119886
119894119895119896 such that 119886
119894119895119896ge
0 for 119886119894119895119896isin 119866 and 119886
119894119895119896= 0 for 119886
119894119895119896isin 119867 using dynamical
programming
Acknowledgments
The author is grateful to Professor Y Dinitz and Professor GGutin for their attention to this work and valuable commentsAlso the author is grateful to the anonymous referee of thisjournal for the valuable comments and suggestions
References
[1] R A Brualdi and J Csima ldquoOn the plane term rank of a threedimensional matrixrdquo Proceedings of the AmericanMathematicalSociety vol 54 no 1 pp 471ndash473 1976
[2] J Csima ldquoOn the plane term rank of three dimensional matri-cesrdquo Discrete Mathematics vol 28 no 2 pp 147ndash152 1979
[3] D K Kim Y A Kim and K Park ldquoGeneralizations of suffixarrays to multi-dimensional matricesrdquo Theoretical ComputerScience vol 302 no 1ndash3 pp 401ndash416 2003
[4] Y Kawamura ldquoCubic matrices generalized spin algebra anduncertainty relationrdquo Progress ofTheoretical Physics vol 110 no3 pp 579ndash587 2003
[5] R E Burkard R Rudolf and G J Woeginger ldquoThree-dimen-sional axial assignment problems with decomposable cost coef-ficientsrdquoDiscrete Applied Mathematics vol 65 no 1ndash3 pp 123ndash139 1996
[6] W P Pierskalla ldquoThe multidimensional assignment problemrdquoOperations Research vol 16 no 2 pp 422ndash431 1968
[7] M R Garey and D C Johnson Computer and Intractability AGuide To the Theory of NP-Completeness W H Freeman SanFrancisco Calif USA 1979
[8] W W Bein P Brucker J K Park and P K Pathak ldquoA Mongeproperty for the d-dimensional transportation problemrdquo Dis-crete Applied Mathematics vol 58 no 2 pp 97ndash109 1995
[9] D A Blokh ldquoAssignment problem for three-index Jacobimatri-cesrdquoAutomation and Remote Control vol 50 no 5 pp 672ndash6771989
[10] Y Crama and F C R Spieksma ldquoApproximation algorithms forthree-dimensional assignment problems with triangle inequal-itiesrdquo European Journal of Operational Research vol 60 no 3pp 273ndash279 1992
[11] F C R Spieksma and G J Woeginger ldquoGeometric three-dimensional assignment problemsrdquo European Journal of Oper-ational Research vol 91 no 3 pp 611ndash618 1996
[12] E Balas and M J Saltzman ldquoAn algorithm for the three-indexassignment problemrdquo Operations Research vol 39 no 1 pp150ndash161 1991
[13] G Gutin B Goldengorin and J Huang ldquoWorst case analysisof Max-Regret greedy and other heuristics for multidimen-sional assignment and traveling salesman problemsrdquo Journal ofHeuristics vol 14 no 2 pp 169ndash181 2008
[14] B-J Kim W L Hightower P M Hahn Y-R Zhu and L SunldquoLower bounds for the axial three-index assignment problemrdquoEuropean Journal of Operational Research vol 202 no 3 pp654ndash668 2010
[15] D Blokh and G Gutin ldquoMaximizing traveling salesman prob-lem for special matricesrdquo Discrete Applied Mathematics vol 56no 1 pp 83ndash86 1995
[16] D Blokh and E Levner ldquoAn approximation algorithm withperformance guarantees for the maximum traveling salesmanproblemon specialmatricesrdquoDiscrete AppliedMathematics vol119 no 1-2 pp 139ndash148 2002
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Discrete Mathematics 3
Theorem 1 it is possible to make such replacement of entriessuch that the sum of T set entries will not change and thenumber of 119879 cap 119867 set entries is reduced
Theorem 5 If 119886119894119895119896 isin 119880(119899) then the AP3 optimum for a
1198861015840
119894119895119896 matrix where 1198861015840
119894119895119896= 119886119894119895119896
for 1198861015840119894119895119896isin 119866 and 1198861015840
119894119895119896= 0 for
1198861015840
119894119895119896isin 119867 represents an approximation for the AP3 optimum of
the 119886119894119895119896matrix with a relative error not exceeding 1(3C)
Proof Let 119886119894119895119896 isin 119880(119899) 119898
1is the AP3 optimum for 119886
119894119895119896
119891 119892 isin 119878119899are such that119898
1= sum119899
119894=1119886119894119891(119894)119892(119894)
and the number of119886119894119891(119894)119892(119894)
| 119894 = 1 119899 cap 119867 set entries is no more than [1198993]1198982is the sum of 119886
119894119891(119894)119892(119894)| 119894 = 1 119899 cap 119867 set entries and
1198983= 1198981minus 11989821198981015840 is the AP3 optimum for 1198861015840
119894119895119896
Note that1198983= sum119899
119894=11198861015840
119894119891(119894)119892(119894)and1198981015840 ge 119898
3
Insofar as 119886119894119895119896ge 1198861015840
119894119895119896 119894 119895 119896 = 1 119899 then 119898
1ge 1198981015840
Hence1198983le 1198981015840
le 1198981and (119898
1minus 1198981015840
)1198981le (1198981minus 1198983)1198981=
11989821198981
Since1198982le (13)119899sdotmax
119886119894119895119896isin119867119886119894119895119896
and1198981ge 119899sdotmin
119886119894119895119896isin119866119886119894119895119896
then 11989821198981le 1(3119862) Hence (119898
1minus 1198981015840
)1198981le 1(3119862) and
the relative error of1198981015840 taken as the AP3 approximate solutionfor the 119886
119894119895119896matrix is no more than 1(3119862)
Remark 6 An exact119874(119899)-algorithmofAP3 solution has beenconstructed [9] for 119899 times 119899 times 119899-matrices 119886
119894119895119896 such that 119886
119894119895119896ge
0 for 119886119894119895119896isin 119866 and 119886
119894119895119896= 0 for 119886
119894119895119896isin 119867 using dynamical
programming
Acknowledgments
The author is grateful to Professor Y Dinitz and Professor GGutin for their attention to this work and valuable commentsAlso the author is grateful to the anonymous referee of thisjournal for the valuable comments and suggestions
References
[1] R A Brualdi and J Csima ldquoOn the plane term rank of a threedimensional matrixrdquo Proceedings of the AmericanMathematicalSociety vol 54 no 1 pp 471ndash473 1976
[2] J Csima ldquoOn the plane term rank of three dimensional matri-cesrdquo Discrete Mathematics vol 28 no 2 pp 147ndash152 1979
[3] D K Kim Y A Kim and K Park ldquoGeneralizations of suffixarrays to multi-dimensional matricesrdquo Theoretical ComputerScience vol 302 no 1ndash3 pp 401ndash416 2003
[4] Y Kawamura ldquoCubic matrices generalized spin algebra anduncertainty relationrdquo Progress ofTheoretical Physics vol 110 no3 pp 579ndash587 2003
[5] R E Burkard R Rudolf and G J Woeginger ldquoThree-dimen-sional axial assignment problems with decomposable cost coef-ficientsrdquoDiscrete Applied Mathematics vol 65 no 1ndash3 pp 123ndash139 1996
[6] W P Pierskalla ldquoThe multidimensional assignment problemrdquoOperations Research vol 16 no 2 pp 422ndash431 1968
[7] M R Garey and D C Johnson Computer and Intractability AGuide To the Theory of NP-Completeness W H Freeman SanFrancisco Calif USA 1979
[8] W W Bein P Brucker J K Park and P K Pathak ldquoA Mongeproperty for the d-dimensional transportation problemrdquo Dis-crete Applied Mathematics vol 58 no 2 pp 97ndash109 1995
[9] D A Blokh ldquoAssignment problem for three-index Jacobimatri-cesrdquoAutomation and Remote Control vol 50 no 5 pp 672ndash6771989
[10] Y Crama and F C R Spieksma ldquoApproximation algorithms forthree-dimensional assignment problems with triangle inequal-itiesrdquo European Journal of Operational Research vol 60 no 3pp 273ndash279 1992
[11] F C R Spieksma and G J Woeginger ldquoGeometric three-dimensional assignment problemsrdquo European Journal of Oper-ational Research vol 91 no 3 pp 611ndash618 1996
[12] E Balas and M J Saltzman ldquoAn algorithm for the three-indexassignment problemrdquo Operations Research vol 39 no 1 pp150ndash161 1991
[13] G Gutin B Goldengorin and J Huang ldquoWorst case analysisof Max-Regret greedy and other heuristics for multidimen-sional assignment and traveling salesman problemsrdquo Journal ofHeuristics vol 14 no 2 pp 169ndash181 2008
[14] B-J Kim W L Hightower P M Hahn Y-R Zhu and L SunldquoLower bounds for the axial three-index assignment problemrdquoEuropean Journal of Operational Research vol 202 no 3 pp654ndash668 2010
[15] D Blokh and G Gutin ldquoMaximizing traveling salesman prob-lem for special matricesrdquo Discrete Applied Mathematics vol 56no 1 pp 83ndash86 1995
[16] D Blokh and E Levner ldquoAn approximation algorithm withperformance guarantees for the maximum traveling salesmanproblemon specialmatricesrdquoDiscrete AppliedMathematics vol119 no 1-2 pp 139ndash148 2002
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of