Brian McFee

Circling back on this, i benchmarked the beat tracker before and after the proposed bug fix in the dp algorithm, using them isophonics set. Tldr, the proposed change results in...

> Will have to investigate this further. Digging back in now, and it seems like the score drop is probably coming from changes in the trimming behavior. With trimming enabled...

> With trimming disabled, it continues to estimate beats to the end of the track where the old implementation did not. Aha, I see what's happening here. The trimming logic...

> I suppose for the sake of consistency within a minor revision (0.10.1 → 0.10.2) we probably should preserve this behavior, but it might make sense to disable that going...

> At this point, I think this PR is done. Ok, done except for a little bit of test polish and some comment cleanup. But code-wise, this is ready for...

I recognize that this PR is probably un-reviewable. :shrug: That said, if anyone wants to do a skim of the documentation sections, it would be most helpful. I think the...

For funsies, I redid the audio example above using a snare drum sample and put together an advanced demo notebook to show how it works. Punchline image:  Is this...

Just a bit more detail on the example, here are the accuracy results from `mir_eval`: | | F-measure | Cemgil | Cemgil Best Metric Level | Goto | P-score |...

Thanks @lostanlen ! I've updated the text mostly taking your suggestions (with some minor tweaks here and there). Exception messages have also been expanded per your suggestion, though the phrasing...

Doc CI failure is due to a temporary outage of the scipy procedings archive, nothing to be done from our side.