Roadmap
Next up
Features
- Improve handling of multi-fidelity for large scale (slurm script modification)
- Evaluate and maybe improve ease-of-use of NePS and DDP etc.
- Optimize dependencies
- Improved examples
Fixes
- Acq search mutation for HPs potentially only mutates 1 parameter
ignore_errors
should work seamlessly with all optimizers
Refactoring
- Rename: run_pipeline = evaluate_pipeline | evaluate_pipeline_error | compute_pipeline_error | train_and_evaluate
- Rename: loss = validation_error | error | pipeline_error
- Rename: XParameter = XSpace or just X?
- Rename: default-x to prior-x
- Rename: Use max_cost_total everywhere instead of budget
Documentation
- Keep citations doc up to date
Tests
- Regression tests to run on each push
Before 1.0.0 version
Features
- Generate pdf plot after each evaluation
- Finegrained control over user prior
- Print search space upon run
- Utility to generate code for best architecture
- Core algorithmic feature set (research)
Fixes
Refactoring
- Improve neps.optimizers:
- Maintained vs unmaintained optimizers
- Remove unnecessary / broken optimizers
- Merge GP and hierarchical GP
- Break up search space and config aspect
Documentation
After 1.0.0
Features
- Utility neps.clean to manage existing run results
- Collect data optionally via phone-home to webserver
Documentation