boax.acquisitions.surrogates.single_task_gaussian_process

boax.acquisitions.surrogates.single_task_gaussian_process#

boax.acquisitions.surrogates.single_task_gaussian_process(bounds, fit_learning_rate=0.01, fit_num_steps=500, optimizer_num_raw_samples=512, optimizer_num_restarts=10)#

The single task gaussian process surrogate model.

Has strict priors on its paramters.

Example

>>> surrogate = single_task_gaussian_process(bounds, learning_rate, num_steps)
>>> params = surrogate.init()
>>> fitted_params = surrogate.update(params, observation_index_points, observations)
>>> prior = surrogate.prior(params)
>>> posterior = surrogate.update(params, observation_index_points, observations)
>>> best = surrogate.best(posterior)
Parameters:

  • bounds (Union[Array, ndarray, bool, number]) – The domain bounds.

  • learning_rate – The learning for fitting the parameters during parameter update.

  • num_steps – The number of steps for fitting the parameters during parameter update.

Return type:

Surrogate[dict, MultivariateNormal]

Returns;

The corresponding Surrogate.