Alexander Held
Alexander Held
This tracks possible / required changes to `cabinetry` for the `pyhf` 0.8 release. **Required:** - [ ] https://github.com/scikit-hep/cabinetry/pull/385 Optional: - [ ] better `strategy=None` handling (see https://github.com/scikit-hep/cabinetry/pull/330 and https://github.com/scikit-hep/pyhf/issues/1785, resolved...
Allow repeating fits with all parameters but POI(s) set to best-fit points to extract statistical uncertainty. - other normalization factors: could be made configurable - may be useful to set...
This collects information regarding changes in the `cabinetry` visualization API, and is a follow-up to #251. - #250 made `matplotlib` a core dependency and refactored the plotting code. - #264...
With `dask-awkward` and integration with `hist`, it seems conceptually possible to make the template histogram creation fully Dask-driven. This could result in two significant improvements: - parallelization of template histogram...
This is a partial follow-up to #315. @ekauffma found that the yield uncertainty calculation can use a large amount of memory. This ultimately is due to using `awkward` instead of...
Adding functionality to re-bin during template post-processing. _to-do:_ - decide whether to support index- or coordinate-based rebinning (or both?) - understand typing of histogram re-assignment - converge on config format...
Add functionality to `model_utils.prediction` that allows bypassing uncertainty calculations. This could be done directly via `pyhf`'s `model.expected_data`, but wrapping it allows using the same consistent API, which might be useful.
Provide the possibility to re-bin histograms, which probably should happen at the post-processing step. The rebinning could be done via `hist` and the syntax should presumably follow [UHI](https://uhi.readthedocs.io/en/latest/). Generic rebinnings...
This is a follow-up of #219. #289 implements method 4) of #219: it extends the config schema to support reading histogram inputs (as an alternative to ntuples), and also adds...
> I had some perf suggestions after this popped up in my notifications: > > Pre-convert the parameters, and use `np.ufunc.at` to avoid a temporary > ```python > # calculate...