Function matchpairs does not work in MATLAB R2023a when I use a sparse cost matrix. Please help me out ASAP!

You can run this simple code and see my problem. Am I doing something wrong?

% testing matchpairs ----------------------------------------------------------------
clear;clc
% --- using matrix in usual form (not sparse) --- WORKS
M = 99*ones(16,16);
M(1,1)   = 2;
M(2,1)   = 1;
M(1,2)   = 1;
M(3,2)   = 2;
M(2,3)   = 2;
M(4,3)   = 1;
M(6,3)   = 4;
M(3,4)   = 1;
M(4,4)   = 2;
M(7,4)   = 4;
M(6,5)   = 2;
M(8,5)   = 1;
M(3,6)   = 4;
M(5,6)   = 2;
M(7,6)   = 1;
M(4,7)   = 4;
M(6,7)   = 1;
M(10,7)  = 2;
M(5,8)   = 1;
M(9,8)   = 2;
M(8,9)   = 2;
M(12,9)  = 1;
M(7,10)  = 2;
M(11,10) = 1;
M(13,10) = 4;
M(10,11) = 1;
M(12,11) = 2;
M(14,11) = 4;
M(9,12)  = 1;
M(11,12) = 2;
M(10,13) = 4;
M(13,13) = 2;
M(14,13) = 1;
M(11,14) = 4;
M(13,14) = 1;
M(15,14) = 2;
M(14,15) = 2;
M(16,15) = 1;
M(15,16) = 1;
M(16,16) = 2;
[matchings, unassignedRows, unassignedCols] = matchpairs(M,12345) 

matchings = 16x2

2 1 1 2 4 3 3 4 8 5 7 6 6 7 5 8 12 9 11 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

unassignedRows =

  0x1 empty double column vector


unassignedCols =

  0x1 empty double column vector

OUTPUT IS CORRECT:

% --- using sparse matrix --- DOES NOT WORK
Ms = sparse(16,16);
Ms(1,1)   = 2;
Ms(2,1)   = 1;
Ms(1,2)   = 1;
Ms(3,2)   = 2;
Ms(2,3)   = 2;
Ms(4,3)   = 1;
Ms(6,3)   = 4;
Ms(3,4)   = 1;
Ms(4,4)   = 2;
Ms(7,4)   = 4;
Ms(6,5)   = 2;
Ms(8,5)   = 1;
Ms(3,6)   = 4;
Ms(5,6)   = 2;
Ms(7,6)   = 1;
Ms(4,7)   = 4;
Ms(6,7)   = 1;
Ms(10,7)  = 2;
Ms(5,8)   = 1;
Ms(9,8)   = 2;
Ms(8,9)   = 2;
Ms(12,9)  = 1;
Ms(7,10)  = 2;
Ms(11,10) = 1;
Ms(13,10) = 4;
Ms(10,11) = 1;
Ms(12,11) = 2;
Ms(14,11) = 4;
Ms(9,12)  = 1;
Ms(11,12) = 2;
Ms(10,13) = 4;
Ms(13,13) = 2;
Ms(14,13) = 1;
Ms(11,14) = 4;
Ms(13,14) = 1;
Ms(15,14) = 2;
Ms(14,15) = 2;
Ms(16,15) = 1;
Ms(15,16) = 1;
Ms(16,16) = 2;
[matchings, unassignedRows, unassignedCols] = matchpairs(Ms,-1)

matchings =

  0x2 empty double matrix

unassignedRows = 16x1

1 2 3 4 5 6 7 8 9 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

unassignedCols = 16x1

1 2 3 4 5 6 7 8 9 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Answer 1

[Bruno Luong]

Your sparse matrix filled default cost wit 0 not 99. Then your expectation on that the result is identocal for different cost matrices (regardless data structure using) is simply plain wrong.

If you fill sparse or dense matrix with the same values it give identical outputs:

% testing matchpairs ----------------------------------------------------------------
clear;clc
% --- using matrix in usual form (not sparse) --- WORKS
M = 99*ones(16,16);
M(1,1)   = 2;
M(2,1)   = 1;
M(1,2)   = 1;
M(3,2)   = 2;
M(2,3)   = 2;
M(4,3)   = 1;
M(6,3)   = 4;
M(3,4)   = 1;
M(4,4)   = 2;
M(7,4)   = 4;
M(6,5)   = 2;
M(8,5)   = 1;
M(3,6)   = 4;
M(5,6)   = 2;
M(7,6)   = 1;
M(4,7)   = 4;
M(6,7)   = 1;
M(10,7)  = 2;
M(5,8)   = 1;
M(9,8)   = 2;
M(8,9)   = 2;
M(12,9)  = 1;
M(7,10)  = 2;
M(11,10) = 1;
M(13,10) = 4;
M(10,11) = 1;
M(12,11) = 2;
M(14,11) = 4;
M(9,12)  = 1;
M(11,12) = 2;
M(10,13) = 4;
M(13,13) = 2;
M(14,13) = 1;
M(11,14) = 4;
M(13,14) = 1;
M(15,14) = 2;
M(14,15) = 2;
M(16,15) = 1;
M(15,16) = 1;
M(16,16) = 2;
[matchings, unassignedRows, unassignedCols] = matchpairs(M,12345)

matchings = 16x2

2 1 1 2 4 3 3 4 8 5 7 6 6 7 5 8 12 9 11 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

unassignedRows =

  0x1 empty double column vector


unassignedCols =

  0x1 empty double column vector

matchings

matchings = 16x2

2 1 1 2 4 3 3 4 8 5 7 6 6 7 5 8 12 9 11 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Ms = sparse(M);
[matchings, unassignedRows, unassignedCols] = matchpairs(M,12345);

matchings

matchings = 16x2

2 1 1 2 4 3 3 4 8 5 7 6 6 7 5 8 12 9 11 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Cross check, now both filled with 0s

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% --- using sparse matrix --- DOES NOT WORK
Ms = sparse(16,16);
Ms(1,1)   = 2;
Ms(2,1)   = 1;
Ms(1,2)   = 1;
Ms(3,2)   = 2;
Ms(2,3)   = 2;
Ms(4,3)   = 1;
Ms(6,3)   = 4;
Ms(3,4)   = 1;
Ms(4,4)   = 2;
Ms(7,4)   = 4;
Ms(6,5)   = 2;
Ms(8,5)   = 1;
Ms(3,6)   = 4;
Ms(5,6)   = 2;
Ms(7,6)   = 1;
Ms(4,7)   = 4;
Ms(6,7)   = 1;
Ms(10,7)  = 2;
Ms(5,8)   = 1;
Ms(9,8)   = 2;
Ms(8,9)   = 2;
Ms(12,9)  = 1;
Ms(7,10)  = 2;
Ms(11,10) = 1;
Ms(13,10) = 4;
Ms(10,11) = 1;
Ms(12,11) = 2;
Ms(14,11) = 4;
Ms(9,12)  = 1;
Ms(11,12) = 2;
Ms(10,13) = 4;
Ms(13,13) = 2;
Ms(14,13) = 1;
Ms(11,14) = 4;
Ms(13,14) = 1;
Ms(15,14) = 2;
Ms(14,15) = 2;
Ms(16,15) = 1;
Ms(15,16) = 1;
Ms(16,16) = 2;
[matchings, unassignedRows, unassignedCols] = matchpairs(Ms,-1);
matchings

matchings =

  0x2 empty double matrix

Mf = full(Ms);
[matchings, unassignedRows, unassignedCols] = matchpairs(Mf,-1);
matchings

matchings =

  0x2 empty double matrix

[Bruno Luong]

One way to use sparse and keep filling 0 is to shift the value of the cost matrix

% testing matchpairs ----------------------------------------------------------------


Ms = sparse(16,16);
Ms(1,1)   = 2;
Ms(2,1)   = 1;
Ms(1,2)   = 1;
Ms(3,2)   = 2;
Ms(2,3)   = 2;
Ms(4,3)   = 1;
Ms(6,3)   = 4;
Ms(3,4)   = 1;
Ms(4,4)   = 2;
Ms(7,4)   = 4;
Ms(6,5)   = 2;
Ms(8,5)   = 1;
Ms(3,6)   = 4;
Ms(5,6)   = 2;
Ms(7,6)   = 1;
Ms(4,7)   = 4;
Ms(6,7)   = 1;
Ms(10,7)  = 2;
Ms(5,8)   = 1;
Ms(9,8)   = 2;
Ms(8,9)   = 2;
Ms(12,9)  = 1;
Ms(7,10)  = 2;
Ms(11,10) = 1;
Ms(13,10) = 4;
Ms(10,11) = 1;
Ms(12,11) = 2;
Ms(14,11) = 4;
Ms(9,12)  = 1;
Ms(11,12) = 2;
Ms(10,13) = 4;
Ms(13,13) = 2;
Ms(14,13) = 1;
Ms(11,14) = 4;
Ms(13,14) = 1;
Ms(15,14) = 2;
Ms(14,15) = 2;
Ms(16,15) = 1;
Ms(15,16) = 1;
Ms(16,16) = 2;

% Shift matrix
[I,J,K] = find(Ms);
fillvalue = 99;
Msshift = sparse(I,J,K-fillvalue,16,16); 

[matchings, unassignedRows, unassignedCols] = matchpairs(Msshift,12345);

matchings

matchings = 16x2

2 1 1 2 4 3 3 4 8 5 7 6 6 7 5 8 12 9 11 10

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

[Ulyana Grechaniuk]

Thank you, Bruno !!!