- Christoph Deil
- Nov 18, 2019
- Gammapy coding sprint in Granda
- We introduced
gammapy.modeling
andgammapy.modeling.models
. See docs PIG 16 - PIG 14 - Uncertainty estimation
- gammapy.modeling project board: https://github.com/gammapy/gammapy/projects/7
Things we should try to do for v0.15. I put an estimate of how long it could take. (about as reliable as astronomical distance estimates)
- Write documentation (
docs/modeling/index.rst
,tutorials/models.ipynb
,tutorials/modeling.ipynb
) - Implement PIG 14 - clean up spectral model evaluation, error propagation (2 days)
- Polish model registry and factory (GH 2387) (1 day)
- Clean up model names (GH 2353)
(
lon_0
->lon
,reference
->energy_ref
,lambda_
->alpha
, ...) Needs proposal and decision. Do next week? (1-2 days) - API to interact with parameter errors (GH 1398) (1 day)
- Improve MCMC sampling code GH 2304
Would like to have for v1.0, but probably not a blocker?
- Improve model evaluate and
__call__
API (GH 2443) (needs design proposal, ~ 1 week, defer to 2020?) - Add optional
SkyModel.temporal_model
(2 hours) and make evaluate work properly (1-2 days) - Make SkyDiffuseCube a
SpatialModel
subclass? - Remove parameter scaling? (GH 2434) (1-2 days)
- LogGaussianSpectralModel mean position (GH 2436) (3 hours)
- Dedicated effort on Gammapy modeling 2.0? Hire Tom?
- Goal: 10x to 1000x faster, depending on model and machine
- Issues, solutions, tasks, priorities, contributors?