2

I'm looking to implement spatially correlated large-area errors similar to this paper, where a local patch of qubits experience random $I$,$X$, $Y$, $Z$ errors (simultaneously) with some probability $p(r)$, where $r$ is related to the size of the patch.

I can employ the CORRELATED_ERROR and ELSE_CORRELATED_ERROR instructions to implement this, but because I have to specify each Pauli product, this seemingly becomes intractable as the number of Pauli products grows as $4^N$ for an $N$-qubit error.

Is there some functionality of Stim or a workaround that I'm overlooking that would make this implementable? Or would I have to resort stim.FlipSimulator or stim.TableauSimulator?

FDGod
  • 2,278
  • 2
  • 4
  • 29
jfids
  • 47
  • 5

1 Answers1

1

You should use stim.FlipSimulator or stim.TableauSimulator for error models with huge numbers of terms. The circuits explicitly list every term, which is intractable when you have a million of them.

It may still be useful to produce a circuit, but with a simplified error model, in order to configure your decoder (unless it also computes terms on-demand).

Craig Gidney
  • 36,389
  • 1
  • 29
  • 95