Required prerequisites
Motivation
I am learning BOAT through the examples/data_hyper_cleaning/test_data_hyper_cleaning.py example. The example is useful because it includes many bilevel optimization operator combinations, such as NGD, GDA, DI, DM, CG, RAD, FD, NS, IGA, VSO, VFO, MESO, and PGDO.
However, as a new user, I found it difficult to quickly understand which combinations belong to GM-OL, NA-OL, and FO-OL, and which combinations are recommended for beginners to run first.
This is related to the data hyper-cleaning use case. Since BOAT is configuration-heavy, a beginner-friendly cheatsheet would reduce confusion and help new users run the examples more smoothly.
Proposed solution
I suggest adding a short cheatsheet to the documentation or to the examples/data_hyper_cleaning README.
The cheatsheet could include:
- Recommended beginner combinations, such as NGD + CG, GDA + CG, and NGD + RAD.
- A table explaining which operators belong to GM-OL, NA-OL, and FO-OL.
- Common invalid or confusing combinations that users should avoid.
- A short explanation of what each operator combination is intended to test.
- A minimal command for running one stable data hyper-cleaning example.
This would make the data hyper-cleaning example easier to understand and reduce trial-and-error configuration mistakes.
Alternatives considered
No response
Additional context
No response
Required prerequisites
Motivation
I am learning BOAT through the examples/data_hyper_cleaning/test_data_hyper_cleaning.py example. The example is useful because it includes many bilevel optimization operator combinations, such as NGD, GDA, DI, DM, CG, RAD, FD, NS, IGA, VSO, VFO, MESO, and PGDO.
However, as a new user, I found it difficult to quickly understand which combinations belong to GM-OL, NA-OL, and FO-OL, and which combinations are recommended for beginners to run first.
This is related to the data hyper-cleaning use case. Since BOAT is configuration-heavy, a beginner-friendly cheatsheet would reduce confusion and help new users run the examples more smoothly.
Proposed solution
I suggest adding a short cheatsheet to the documentation or to the examples/data_hyper_cleaning README.
The cheatsheet could include:
This would make the data hyper-cleaning example easier to understand and reduce trial-and-error configuration mistakes.
Alternatives considered
No response
Additional context
No response