Nevergrad

class Nevergrad(_opt_id=None, _signal_pipe=None, _results_queue=None, _pause_flag=None, _is_log_detailed=False, _workers=1, _backend='threads', optimizer='TBPSA', zero=0, warn=False, **kwargs)[source]

Bases: BaseOptimizer

Provides access to the optimizers available through the nevergrad package. Tested with v.0.3.2.

Parameters:

_opt_id _signal_pipe _results_queue _pause_flag _is_log_detailed _workers _backend

See BaseOptimizer.

optimizer

String key to the desired optimizer. See nevergrad documentation for a list of available algorithms.

zero

Will stop the optimization when this cost function value is reached.

warn

If True, suppresses all nevergrad warnings.

**kwargs

Extra initialisation arguments passed to the optimizer.

Notes:

Important

It is crucial that you check for possible search space settings of the requested optimizer and adjust them through the kwargs argument accordingly. By default, ParAMS will return an infinite loss function value if any of the parameters is out of bounds. Certain algorithms (especially if the dimensionality is high) might not be able to adjust to the bounded search space, resulting in all evaluations producing float(inf), and a failed optimization.