1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
|
// pour tester condestsp
//
// on utilise des petites matrices pour verifier
nb_tests = 20;
n = 100;
K1 = zeros(nb_tests,3);
messagebox(["A small test for the condestsp" ;...
"function : condestsp gives an" ;...
"estimate of the condition number K1";...
"in 1-norm of a real sparse matrix A:";...
" " ;...
" K1 = || A ||_1 || A^(-1) ||_1 " ;...
" " ;...
"without explicitly computing the" ;...
"inverse of A. condestsp uses a" ;...
"factorization given by umf_lufact" ;...
"but if you have already computed" ;...
"this one it is recommended to give" ;...
"the pointer onto the factorization.";...
" ";...
"The test consists in forming small ";...
"sparse matrices (so as to compute ";...
"exactly K1 with norm(inv(full(A)),1))";...
"whose values are choosen from the" ;...
"normal distribution. "],"modal","info");
for k = 1:nb_tests
m = grand(1,1,"uin",1000,2000); // nnz
ij = grand(m,2,"uin",1,100);
v = grand(m,1,"nor",0,1);
A = sparse(ij,v,[n n]);
Lup = umf_lufact(A);
K1(k,1) = condestsp(A,Lup);
K1(k,2) = condestsp(A,Lup,5);
K1(k,3) = norm(A,1)*norm(inv(full(A)),1);
umf_ludel(Lup)
end
str2 = "------------";
str3 = "condest t=2";
str4 = "condest t=5";
str5 = " K1 exact ";
mprintf(" +-%s-+-%s-+-%s-+ \n",str2,str2,str2)
mprintf(" | %s | %s | %s | \n",str3,str4,str5)
mprintf(" +-%s-+-%s-+-%s-+ \n",str2,str2,str2)
for k=1:nb_tests
mprintf(" | %e | %e | %e | \n",K1(k,1:3))
end
mprintf(" +-%s-+-%s-+-%s-+ \n",str2,str2,str2)
|
