# Thread: ask for C Numerical Libraries about matrix of multiply and subtraction

1. ## ask for C Numerical Libraries about matrix of multiply and subtraction

may i ask for how to solve linear equation by using sparse matrix in IMSL C Numerical Libraries

Mathematical formulas just like :
A*x=B*b+C*c , solve for x=?
(A=B=C=10*10 matrix) (x=b=c=10*1 matrix)

how can i get the inverse of A by using sparse system
then the solution x=A-1 *B*b+A-1 *C*c can be obtained

thank you for your attention very much  Reply With Quote

2. Originally Posted by kll94530136 may i ask for how to solve linear equation by using sparse matrix in IMSL C Numerical Libraries

Mathematical formulas just like :
A*x=B*b+C*c , solve for x=?
(A=B=C=10*10 matrix) (x=b=c=10*1 matrix)
B*b + C*c is also a vector, and can be computed as such before solving the simultaneous equations. Let us call the result q. Then, the equation set to solve is A*x = q.
how can i get the inverse of A by using sparse system
then the solution x=A-1 *B*b+A-1 *C*c can be obtained
Many mathematical expressions that look reasonable, such as your result for x, fall into the category of "bad idea" in a numerical computation. One such is computing the inverse of a matrix as a stepping-stone to the solution of simultaneous equations. Sparse solvers will let you obtain the solution of A x = q without having to compute the inverse.

Which solver should you use? That depends on whether A is banded or not. Besides that, if A is only 10 X 10 there is not enough inducement to exploit sparsity.  Reply With Quote

3. Originally Posted by mecej4 B*b + C*c is also a vector, and can be computed as such before solving the simultaneous equations. Let us call the result q. Then, the equation set to solve is A*x = q.

thanks
i got it

Many mathematical expressions that look reasonable, such as your result for x, fall into the category of "bad idea" in a numerical computation. One such is computing the inverse of a matrix as a stepping-stone to the solution of simultaneous equations. Sparse solvers will let you obtain the solution of A x = q without having to compute the inverse.

Then i want to ask for a more question
is that sparse matrix can only solve for
square matrix??
if i want B*b (B=50*49 matrix , b=49*1 matrix )

imsl_f_mat_mul_rect_coordinate can realize it?
if it can , can you teach me how to do
thank you very much

Which solver should you use? That depends on whether A is banded or not. Besides that, for a if A is only 10 X 10 there is not enough inducement to exploit sparsity.
My matrix is very huge,just like

A*x=B*b+C*c
[2110*2110][2110*1]=[2110*1000][1000*1]+[2110*2110][2110*1]

and the matrix form is not banded
so it can be solved for only sparse in appropriately

oh ,there is an another question about output
if i want to let the answers be stored in a .txt format
how can i do it?
the User Guide is written in
imsl_output_file
but i don't understand very much  Reply With Quote

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•