ArviZ 2019 Roadmap
What stuff do we want in ArviZ?
- Artists
- Linear Counter Factual plots?
- VI support?
Google Summer of Code proposals include details on desired functionality as well https://github.com/arviz-devs/arviz/wiki/GSoC-2019-projects
Currently ArviZ test suite tests plotting and conversion numerous times, and requires all inference libraries to be present to test plotting and statistics code easily. Benefits of InferenceData abstraction would include
- Simpler testing of inference library independent functionality
- Elimination of repeated testing on same plotting code
General architecture is as proposed below
Currently testing requires all inference libraries to be present. For ease of use, and to encourage development, it would be beneficial to allow portions of test code to be run when only one Inference library is present
https://github.com/arviz-devs/arviz/issues/553
ArviZ CI rebuilds entire docker container for every test branch. Caching certain portions would reduce CI time, helping increase development speed and reduce infrastructure load.
https://github.com/arviz-devs/arviz/issues/562
Creation of educational material related to exploratory analysis of Bayesian models with ArviZ. The material will cover topics like data visualization and exploratory analysis of Bayesian models, thus we hope this will help to increase the literacy of the community regarding the theory and good practices related to Bayesian data analysis.
https://github.com/arviz-devs/arviz/wiki/TOC-educational-resources