Interface with popular package estimators - Githubissues

rademacher-p / stats-learn

Python package for statistical learning with a Bayesian focus

https://rademacher-p.github.io/stats-learn/

MIT License

1 stars 0 forks source link

Interface with popular package estimators #9

Closed rademacher-p closed 2 years ago

rademacher-p commented 3 years ago

Learning
- [x] fit, predict, evaluate
- [x] Need support for learners that cannot warm start?
- Logic in MC funcs to pass all previous data
- [ ] set_params
- Need full re-train logic here, as well?
Plotting
- [x] Preserve name attribute?
- [x] Attr space_x used for predict plots
- Make it a func argument instead, deprecate x?!?
- [x] Remove predictor space arg?
- [ ] Avoid tex_params
- Make a registry for str mappings?

Just add additional predictor objects to ignore set_params and tex_params!?!

Packages
- [x] scikit-learn
- [ ] tensorflow?

rademacher-p commented 3 years ago

Ignoring Tensorflow for now. State-preserving parameter setting not needed for desired results. TODO: add params logic in Monte Carlo functions to disallow modifying inappropriate predictor attributes.