Antoine Guillaume
Antoine Guillaume
Summarizing what we discussed Linked to this issue : https://github.com/numba/numba/issues/9348 The issue happens because of cache corruption on numba functions taking other numba function as parameter. Potential fixes in order...
Actions to fix RDST : - [x] Deprecate the distance argument to only use Manhattan, as in the paper. This argument also causes RDST to be a lot slower than...
Re-open and will close with the similarity search related PR
Will take a look tomorrow
Hi Ivan, while memory wise the storage of these parameters would likely not cause too much trouble, I feel like this kind of optional storage should be gated behind a...
One option would be to have the estimator as a mandatory parameter, moving the burden of choosing the correct one depending on the task to the user. Regarding the scoring...
If this is indeed a published method Matthew would be best
> I am struggling to figure out why the deterministic results for RDST/RIST have changed. This is not happening in any other PR. I thought maybe it was due to...
I remember some bug with RDST that only appeared sporadically on macOS, maybe this is kind of the same thing? Don't have a certain answer, only that the changes in...
@lucifer4073 Are you still working on this ? If so do you need any additional inputs?