Budget

Budget

Small: Search a small subset of the hyperparameter space. This will be fast but less optimal.

Medium: Search a medium-sized subset of the hyperparameter space. This balances runtime against optimality.

Large: Search a large subset of the hyperparameter space. This will take longer but yields the most optimal results.

Fine-Tuning

Empirical effective lower bound for tuning one ML-model via Genetic Algorithm is 12, below which there are no noticeable improvements over random search. Therefore, we aim to either tune each model at least 12 times or not at all.