How to decompose errors given a DetectorErrorModel in Stim?

Question

Suppose I'm presented with a detector error model (.dem) file that does not have its error mechanisms decomposed into edge-like error mechanisms (i.e. no ^'s appearing). Is there a way to apply the decomposition directly to the DEM object in the python API?

Answer 1

There's no built-in method to do this.

You could write python code to identify the graphlike errors (the ones with at most two detectors), and then attempt to decompose each non-graphlike errors into the known graphlike errors. Alternatively, if the detectors have coordinate data, you may be able to use that to identify which subgraph each detector is associated with and group the detectors into graphlike pieces that way. A more expensive strategy might be to actually ask a decoder to decode the symptoms of each of the non-graphlike errors, and use the edges it returns as their decompositions.