3
$\begingroup$

I have a 5552 by 5552 sparse matrix:

K = SparseArray[
     {
      Band[{1, 1}] -> 0.2, 
      Band[{2, 1}] -> 0.5, 
      Band[{1, 2}] -> 1., 
      Band[{300, 1}] -> 1., 
      Band[{1000, 1}] -> 1.
     }, {5552, 5552}];

An array is given:

bc = RandomInteger[5552, {542}];

Now my goal is to take each entry of bc and make the corresponding row and column of K zero with [i,i] element as 1.

I am doing it this way:

list1 = Map[{#, All} &, bc];
list2 = Map[{All, #} &, bc];
list3 = Map[{#, #} &, bc];

MapAt[0., K, list1]; // AbsoluteTiming
(* Out: {1.38702, Null} *)

MapAt[0., K, list2]; // AbsoluteTiming
(* Out: {181.601, Null} *)

MapAt[1., K, list3]; // AbsoluteTiming
(* Out: {1.24544, Null} *)

It is taking 181 sec to make rows 0. Is there a faster way to do this?

Improve this question
$\endgroup$
5
  • 4
    $\begingroup$ K is not a good symbol for a variable, it has some internal special meaning. Notice that its color is different from other variables like W or A. $\endgroup$
    – Roman
    Feb 10, 2022 at 8:03
  • 3
    $\begingroup$ you probably meant MapAt[0.&, K, list1] (similarly for other MapAts)? $\endgroup$
    – kglr
    Feb 10, 2022 at 8:21
  • 3
    $\begingroup$ Try k[[bc, ;;]] = 0.; k[[;; , bc]] = 0.; k = ReplacePart[k, Transpose@{bc, bc} -> 1.]; on my computer takes around 0.05 seconds . Also RandomInteger[5552, ...] coud possibly generate 0 element which leads to error in replacing, use {1, 5552} instead. $\endgroup$
    – Ben Izd
    Feb 10, 2022 at 9:03
  • $\begingroup$ Thanks, Ben Izd! It's taking 0.007 sec on my system. $\endgroup$
    – S.B.MD.Khaja Moinuddin
    Feb 10, 2022 at 14:06
  • $\begingroup$ Actually, list2 is unnecessary. Instead of MapAt[0.&, k, list2], which is slow, you could do this: k = Transpose[k]; k = MapAt[0.&, k, list1]; k = Transpose[k];, which is fast. $\endgroup$
    – LouisB
    Feb 16, 2022 at 10:02

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.