view test/diag-perm.tst @ 31119:df030ac26390

uniquetol.m: improve matlab compatibility and add byrows sorting (bug #59850) * /scripts/set/uniquetol.m: improve empty and NaN handling, add sorting to 'byrows' output, ensure ia and ic outputs have column orientation for arrays and cells, verify consistent single class handling, add BISTs for aforementioned cases, and update docstring to note non-complex input requirement.
author Nicholas R. Jankowski <jankowski.nicholas@gmail.com>
date Tue, 05 Jul 2022 15:22:46 -0400
parents 796f54d4ddbf
children 597f3ee61a48
line wrap: on
line source

########################################################################
##
## Copyright (C) 2009-2022 The Octave Project Developers
##
## See the file COPYRIGHT.md in the top-level directory of this
## distribution or <https://octave.org/copyright/>.
##
## This file is part of Octave.
##
## Octave is free software: you can redistribute it and/or modify it
## under the terms of the GNU General Public License as published by
## the Free Software Foundation, either version 3 of the License, or
## (at your option) any later version.
##
## Octave is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Octave; see the file COPYING.  If not, see
## <https://www.gnu.org/licenses/>.
##
########################################################################

########################################
## Permutation matrices

## row permutation
%!test
%! n = 5;
%! A = rand (n);
%! perm = randperm (n);
%! Prow = eye (n) (perm, :);
%! assert (A(perm, :), Prow * A);
%! invperm(perm) = 1:n;
%! assert (Prow \ A, A(invperm, :));
%! assert (Prow' * A, A(invperm, :));

## column permutation
%!test
%! n = 7;
%! A = rand (n);
%! perm = randperm (n);
%! Pcol = eye (n) (:, perm);
%! assert (A(:, perm), A * Pcol);
%! invperm(perm) = 1:n;
%! assert (A / Pcol, A(:, invperm));
%! assert (A * Pcol.', A(:, invperm));

## fall back to a matrix in addition
%!test
%! n = 4;
%! P1 = eye (n) (:, randperm (n));
%! A = zeros (n) + P1;
%! assert (sum (A), full (ones (1, n)));
%! assert (sum (A, 2), full (ones (n, 1)));

## preserve dense matrix structure
%!test
%! n = 7;
%! Pc = eye (n) (:, randperm (n));
%! Pr = eye (n) (randperm (n), :);
%! assert (typeinfo (rand (n) * Pc), "matrix");
%! assert (typeinfo (Pr * rand (n)), "matrix");

## preserve sparse matrix structure
%!test
%! n = 7;
%! Pc = eye (n) (:, randperm (n));
%! Ac = sprand (n-3, n, .5) + I () * sprand (n-3, n, .5);
%! Pr = eye (n) (randperm (n), :);
%! Ar = sprand (n, n+2, .5);
%! assert (typeinfo (Ac * Pc), "sparse complex matrix");
%! assert (full (Ac * Pc), full (Ac) * Pc);
%! assert (full (Ac / Pc), full (Ac) / Pc);
%! assert (typeinfo (Pr * Ar), "sparse matrix");
%! assert (full (Pr * Ar), Pr * full (Ar));
%! assert (full (Pr \ Ar), Pr \ full (Ar));

## structure rules for 1x1 dense / scalar and 1x1 perm
%!test
%! n = 7;
%! P1 = eye (1) (:, [1]);
%! A1 = 1;
%! P = eye (n) (:, randperm (n));
%! A = rand (n-3, n);
%! assert (typeinfo (A * P1), "matrix");
%! assert (full (A * P1), full (A) * P1);
%! assert (typeinfo (P1 * A), "matrix");
%! assert (full (P1 * A), P1 * full (A));
%! assert (typeinfo (A1 * P), "matrix");
%! assert (full (A1 * P), full (A1) * P);
%! assert (typeinfo (P * A1), "matrix");
%! assert (full (P * A1), P * full (A1));

## structure rules for 1x1 sparse and 1x1 perm
%!test
%! n = 7;
%! P1 = eye (1) (:, [1]);
%! A1 = sparse (1, 1, 2);
%! P = eye (n) (:, randperm (n));
%! A = sprand (n-3, n, .5);
%! assert (typeinfo (A * P1), "sparse matrix");
%! assert (full (A * P1), full (A) * P1);
%! assert (typeinfo (P1 * A), "sparse matrix");
%! assert (full (P1 * A), P1 * full (A));
%! assert (typeinfo (A1 * P), "sparse matrix");
%! assert (full (A1 * P), full (A1) * P);
%! assert (typeinfo (P * A1), "sparse matrix");
%! assert (full (P * A1), P * full (A1));

## permuting a matrix with exceptional values does not introduce new ones.
%!test
%! n = 5;
%! pc = randperm (n);
%! Pc = eye (n) (:, pc);
%! pr = randperm (n);
%! Pr = eye (n) (pr, :);
%! A = rand (n);
%! A(n, n-2) = NaN;
%! A(3, 1) = Inf;
%! assert (Pr * A * Pc, A(pr, pc));

## conversion to sparse form
%!test
%! n = 7;
%! P = eye (n) (:, randperm (n));
%! sP = sparse (P);
%! assert (full (sP), full (P));
%! assert (size (find (sP), 1), n);
%! [I, J, V] = find (sP);
%! assert (all (V == 1));

########################################
## Diagonal matrices

## square row scaling
%!test
%! m = 7;
%! n = 11;
%! A = rand (m, n);
%! scalefact = rand (m, 1);
%! Dr = diag (scalefact);
%! assert (Dr * A, repmat (scalefact, 1, n) .* A);
%! assert (Dr \ A, A ./ repmat (scalefact, 1, n));
%! scalefact(m-1) = Inf;
%! Dr(m-1, m-1) = 0;
%! assert (Dr \ A, A ./ repmat (scalefact, 1, n));

## square column scaling
%!test
%! m = 13;
%! n = 11;
%! A = rand (m, n);
%! scalefact = rand (1, n);
%! Dc = diag (scalefact);
%! assert (A * Dc, repmat (scalefact, m, 1) .* A);
%! assert (A / Dc, A ./ repmat (scalefact, m, 1));
%! scalefact(n-1) = Inf;
%! Dc(n-1, n-1) = 0;
%! assert (A / Dc, A ./ repmat (scalefact, m, 1));

## arithmetic
%!test
%! m = 9;
%! n = 7;
%! mn = min (m, n);
%! d1 = rand (mn, 1) + I () * rand (mn, 1);
%! D1 = diag (d1, m, n);
%! d2 = rand (mn, 1);
%! D2 = diag (d2, m, n);
%! D1D2 = D1 + D2;
%! assert (typeinfo (D1D2), "complex diagonal matrix");
%! assert (diag (D1D2), d1 + d2);
%! D1D2 = D2.' * D1;
%! assert (typeinfo (D1D2), "complex diagonal matrix");
%! assert (diag (D1D2), d1 .* d2);

## slicing
## preserving diagonal matrix type is not possible if indices are
## general matrix objects.
%!test
%! m = 13;
%! n = 6;
%! mn = min (m, n);
%! d = rand (mn, 1);
%! D = diag (d, m, n);
%! Dslice = D (1:(m-3), 1:(n-2));
%! if (optimize_range ())
%!   assert (typeinfo (Dslice), "diagonal matrix");
%! else
%!   assert (typeinfo (Dslice), "matrix");
%! endif

## preserve dense matrix structure when scaling
%!assert (typeinfo (rand (8) * (3 * eye (8))), "matrix")
%!assert (typeinfo ((3 * eye (8)) * rand (8)), "matrix")

## preserve sparse matrix structure when scaling
%!assert (typeinfo (sprand (8, 8, .5) * (3 * eye (8))), "sparse matrix")
%!assert (typeinfo (sprand (8, 8, .5) * (3 * eye (8))'), "sparse matrix")
%!assert (typeinfo (((3 + 2 * I ()) * eye (8)) * sprand (8, 8, .5)), "sparse complex matrix")
%!assert (typeinfo (((3 + 2 * I ()) * eye (8))' * sprand (8, 8, .5)), "sparse complex matrix")
%!assert (typeinfo (sprand (8, 8, .5) * ((3 + 2 * I ()) * eye (8)).'), "sparse complex matrix")

## scaling a matrix with exceptional values does not introduce new ones.
%!test
%! n = 6;
%! dr = rand (n, 1);
%! Dr = diag (dr);
%! dc = rand (1, n);
%! Dc = diag (dc);
%! A = rand (n);
%! A(n, n-2) = NaN;
%! A(4, 1) = Inf;
%! assert (Dr * A * Dc, A .* kron (dr, dc), eps);

## sparse inverse row scaling with a zero factor
%!test
%! n = 8;
%! A = sprand (n, n, .5);
%! scalefact = rand (n, 1);
%! Dr = diag (scalefact);
%! scalefact(n-1) = Inf;
%! Dr(n-1, n-1) = 0;
%! assert (full (Dr \ A), full (A) ./ repmat (scalefact, 1, n));

## narrow sparse inverse row scaling
%!test
%! n = 8;
%! A = sprand (n, n, .5);
%! scalefact = rand (n-2, 1);
%! Dr = diag (scalefact, n, n-2);
%! assert (full (Dr \ A), Dr \ full (A));

## sparse inverse column scaling with a zero factor
%!test
%! n = 11;
%! A = sprand (n, n, .5);
%! scalefact = rand (1, n);
%! Dc = diag (scalefact);
%! scalefact(n-1) = Inf;
%! Dc(n-1, n-1) = 0;
%! assert (full (A / Dc), full (A) / Dc);

## short sparse inverse column scaling
%!test
%! n = 7;
%! A = sprand (n, n, .5);
%! scalefact = rand (1, n-2) + I () * rand (1, n-2);
%! Dc = diag (scalefact, n-2, n);
%! assert (full (A / Dc), full (A) / Dc);

## adding sparse and diagonal stays sparse
%!test
%! n = 9;
%! A = sprand (n, n, .5);
%! D = 2 * eye (n);
%! assert (typeinfo (A + D), "sparse matrix");
%! assert (typeinfo (A - D), "sparse matrix");
%! D = D * I () + D;
%! assert (typeinfo (A - D), "sparse complex matrix");
%! A = A * I () + A;
%! assert (typeinfo (D - A), "sparse complex matrix");

## adding sparse and diagonal stays sparse
%!test
%! n = 9;
%! A = sprand (n, n, .5);
%! D = 2 * eye (n);
%! assert (full (A + D), full (A) + D);
%! assert (full (A - D), full (A) - D);
%! D = D * I () + D;
%! assert (full (D + A), D + full (A));
%! A = A * I () + A;
%! A(6, 4) = nan ();
%! assert (full (D - A), D - full (A));

## inverse preserves diagonal structure even for singular matrices (bug #46103)
## but set all the diagonal elements to Inf (bug #56232)
%!test
%! x = diag (1:3);
%! assert (inv (x), diag ([1 1/2 1/3]));
%!warning <matrix singular> A = inv (diag (0:2));
%! assert (A, diag ([Inf Inf Inf]));

## assignment to diagonal elements preserves diagonal structure (bug #36932)
%!test
%! x = diag (1:3);
%! x(1,1) = -1;
%! assert (typeinfo (x), "diagonal matrix");
%! x(3,3) = -1;
%! assert (typeinfo (x), "diagonal matrix");

%!test
%! x = diag (1:3);
%! x(1) = -1;
%! assert (typeinfo (x), "diagonal matrix");
%! x(9) = -1;
%! assert (typeinfo (x), "diagonal matrix");