twinlab.SampleParams

class twinlab.SampleParams(seed=None, observation_noise=False, fidelity=None)

Parameter configuration for sampling from a trained emulator.

Variables:

• seed (Union[int, None], optional) – Specifies the seed used by the random number generator to generate a set of samples. Setting this to an integer is useful for the reproducibility of results. The default value is None, which means the seed is randomly generated each time.

• observation_noise (bool) – Whether or not to include the noise term in the standard deviation of the samples generated. Setting this to False can be a good idea if the training data is noisy but the underlying trend of the trained model is smooth. In this case, the samples would look smooth and would model the underlying trend well. Setting this to True can be a good idea to visualise how noisy are the samples from the emulator, which is a consequence of the noise in the observations. The default value is False.

• fidelity (Union[str, None], optional) – Fidelity information to be provided if the model is a multi-fidelity model (estimator_type="multi_fidelity_gp" in EstimatorParams). This must be a the name of the column in the dataset that corresponds to the fidelity parameter. The default value is None, which is appropriate for most trained emulators.

__init__(seed=None, observation_noise=False, fidelity=None)

Methods

__init__([seed, observation_noise, fidelity])
unpack_parameters()