intelligence.zenith_tune.strategies.random

Random sampling strategy for hyperparameter optimization.

RandomStrategy Objects

@StrategyRegistry.register("random")
class RandomStrategy(TuningStrategy)

Random sampling strategy.

This strategy samples parameters uniformly at random from the search space. For float parameters with log=True, sampling is done in log space.

Runs indefinitely until eval_fn raises BudgetExhausted.

Example:

strategy = RandomStrategy(seed=42) strategy.optimize(search_space, direction, eval_fn)

__init__

def __init__(seed: Optional[int] = None) -> None

Initialize the random strategy.

Arguments:

• seed - Random seed for reproducibility.