How usable is it to implement big Matrices in PL/SQL?

Question

Hi there,
I'm implementing neural network Algorithms with Gradient Descend Methods in PL/SQL. The time for running the Algorithm (Backpropagation of Error) is very large - I still did my best for more Efficiency.
Now I'm thinking about implementing another Algorithm which is called 'Levenberg Marquardt' and that is said to be much more faster. But this Algorithm handles with very large Matrices, invertes and transposes them, needs to create a Jacobian Matrix -> has a big demand for memory
could anyone tell me how efficient it is to implement things like that in PL/SQL? Is there any use in trying it? Perhaps anybody who still worked with something like this?
Thanks in advance
~Mel

Answer 1

Well it's certainly possible in PL/SQL if that's the way you want to implement it...
SQL> ed
Wrote file afiedt.buf
  1  declare
  2    type t_x is table of number index by pls_integer;
  3    type t_y is table of t_x index by pls_integer;
  4    v_xy t_y;
  5  begin
  6    for y in 1..700
  7    loop
  8      for x in 1..700
  9      loop
10        v_xy(y)(x) := ((y-1)*700)+x;
11      end loop;
12    end loop;
13    dbms_output.put_line('y size: '||v_xy.count);
14    dbms_output.put_line('(123,567) = '||v_xy(123)(567));
15    dbms_output.put_line('(700,700) = '||v_xy(700)(700));
16* end;
SQL> /
y size: 700
(123,567) = 85967
(700,700) = 490000
PL/SQL procedure successfully completed.
SQL>