Skip to content

[ENH] Pycatch22 has different results between running via Python and C #1937

Closed
@Moonzyyy

Description

@Moonzyyy

Describe the bug

This is not a bug, but rather an interesting issue with pycatch22.

I recently worked with updating aeon's catch22 to the new C version of pycatch22. The same results were given when comparing the updated catch22 to pycatch22 on the 'italy power demand ' dataset, of course with some minor value differences due to the difference in python and C calculations.

However, testing it on another dataset was interesting as the results were different. I double checked our implementation to the C native of pycatch22 and deemed there was no difference. I tried checking why this was happening. The test I done were this:

  • Grab an array from a dataset where a feature differs between aeon catch22 and pycatch22.
  • Test the array on a python script which import and calls pycatch22 (not aeon), only need to call the feature which gave a discrepancy in the result .
  • Locally test pycatch22 within their C implementation, again just need to check the one feature.

Screenshots are provided to help, currently our version works exactly like pycatch22 C implementation. I have looked through the python wrapper within pycatch22, and see no reason why the change in result occurs.

I already created an issue on their github which can be found here: DynamicsAndNeuralSystems/pycatch22#32 but if anyone is interested to check the difference I'd be happy to discuss it.

Steps/Code to reproduce the bug

No response

Expected results

image

Actual results

image

Versions

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions