Acquisition Functions#

pybads.acquisition_functions.acq_fcn_lcb.acq_fcn_lcb(xi, func_count: int, gp: GP, sqrt_beta=None)[source]#

Lower confidence bound (LCB) of the GP at the points xi.

The acquisition function of the search and of the poll: the GP’s mean at each point less sqrt_beta times its standard deviation.

Parameters:

xinp.ndarray: The points, of shape (N, D).
func_countint: The number of function evaluations so far.
gpgpyreg.GP: The Gaussian process.
sqrt_betaNone, callable or float, optional: The LCB parameter, the multiplier of the GP’s standard deviation. If None, the schedule of Srinivas et al. (2010), with an empirical correction; a callable is called as sqrt_beta(t, D), with t = func_count + 1, and returns a positive finite real number; otherwise a positive finite real number (a Python or NumPy scalar, or an array of one element).

Returns:

Raises:

ValueError: If sqrt_beta is none of the values above, or a callable returns another value.

pybads.acquisition_functions.acq_fcn_lcb.check_sqrt_beta(sqrt_beta, name='sqrt_beta')[source]#

Check the LCB parameter sqrt_beta of acq_fcn_lcb.

Parameters:

sqrt_betaobject: The value to check: None, a callable, or a positive finite real number (a Python or NumPy scalar, or an array of one element).
namestr, optional: What the error message calls sqrt_beta.

Raises:

Acquisition Functions