These algorithms are well-known practical meta-learning GP-BO methods.
As they will be very useful for benchmarking purposes, I would like to have them in OptunaHub.
At least, I would like it to be a single-objective version.
Optionally, it would be great if we could remove the BoTorch dependency using the _gp module in Optuna.
Motivation
I want a single-objective version of RGPE and TSTR from this repository.
These algorithms are well-known practical meta-learning GP-BO methods. As they will be very useful for benchmarking purposes, I would like to have them in OptunaHub.
Description
The implementation here is a multi-objective version and uses BoTorch.
At least, I would like it to be a single-objective version. Optionally, it would be great if we could remove the BoTorch dependency using the
_gpmodule in Optuna.Alternatives (optional)
No response
Additional context (optional)
No response