boax.acquisitions.q_upper_confidence_bound

boax.acquisitions.q_upper_confidence_bound#

boax.acquisitions.q_upper_confidence_bound(bounds, batch_size, sampler, sample_axis=(0,), beta=2.0, num_raw_samples=512, num_restarts=10)#

MC-based batch Upper Confidence Bound acquisition function.

qUCB = E(max(mean + |y_tilde - mean|)),

where y_tilde ~ N(mean(x), beta * pi/2 * cov(x)) and f(x) ~ N(mean(x), cov(x)).

Example

>>> acqf = q_upper_confidence_bound(2.0, model, sampler)
>>> qucb = acqf(index_points)
Parameters:

  • beta (Union[Array, ndarray, bool, number, float, int]) – The mean and covariance trade-off parameter.

  • model – The surrogate model.

  • sampler (Sampler[TypeVar(T)]) – The posterior sampler.

  • sample_axis (Sequence[Union[int, Any]]) – The sample axis.

Return type:

Acquisition

Returns:

The corresponding Acquisition.