neurotools.stats.gridsearch module

Hill-climbing grid search

Grid search hyperparameter optimization

Parameters:

  • pargrid (list of arrays) – A list; Each element is a list of values for a given parameter to search over

  • evaluate (function) –

    Arguments:
    Parameters: Tuple

    Parameters taken from the parameter search grid

    State: List of arrays

    Saves initial conditions (optional, default None)

    Returns:
    state: the inferred model fit, in the form of a list

    of floating-point numpy arrays, to be re-used as initial conditions for subsequent parameters.

    likelihood: float

    Scalar summary of fit quality, higher is better

    info: object

    Anything else you’d like to save

Returns:

  • best – best index into parameter grid

  • pars – values of best parameters

  • results[best] – (state, likelihood, info) at best parameters. info is determined by the third element in the 3-tuple return-value of the evaluate function, passed by the user. state is also user-defined.

  • allresults – all other results