Closed MatthewMiddlehurst closed 1 month ago
this can be a good first issue @MatthewMiddlehurst you agree ?
It requires an understanding of the algorithm and involves comparing to and probably running C code, so probably not.
The alternative would be to switch to pycatch22. https://github.com/aeon-toolkit/aeon/pull/1235
it would be good to profile the two
I still don't really trust it due to previous issues with installation and Windows. If it is proven to be stable then can consider it.
Describe the feature or idea you want to propose
Since its implementation, the numba version of
Catch22
(code here) has become slightly out of date compared to the original implementation.The features which now differ should be updated, and parity with the original output should be verified
Describe your proposed solution
Look at the authors implementation (C code), and update our version of the features in numba to match.
The transformer has a parameter to use this package, so the results can be compared to verify they are the same (or similar).
Describe alternatives you've considered, if relevant
No response
Additional context
No response