baybe.surrogates.utils.catch_constant_targets

baybe.surrogates.utils.catch_constant_targets(cls: type[Surrogate], std_threshold: float = 1e-06)

Make a Surrogate class robustly handle constant training targets.

More specifically, “constant training targets” can mean either of:

• The standard deviation of the training targets is below the given threshold.

• There is only one target and the standard deviation cannot even be computed.

The modified class handles the above cases separately from “regular operation” by resorting to a baybe.surrogates.naive.MeanPredictionSurrogate, which is stored outside the model in a dictionary maintained by this decorator.

Parameters:

• cls (type[Surrogate]) – The baybe.surrogates.base.Surrogate to be augmented.

• std_threshold (float) – The standard deviation threshold below which operation is switched to the alternative model.

Returns:

The modified class.