pele.potentials.ml

class pele.potentials.ml.MLCost

Source

class MLCost(BasePotential):

Cost function to be used for maximum likelihood optimization.

Parameters:

dataarray of floats

The observed data.

log_probfcallable log_probf(data, parameters)

Log of probability function which takes an array of data and an array of parameters and returns a float value

probfcallable probf(data, parameters)

Probability function which takes an array of data and an array of parameters and returns a float value

Examples

To get maximum likelihood estimates of the parameters, do e.g.:

pot = MLCost(observed, probf=gauss) optimizer = LBFGS_CPP(parameters, pot) result = optimizer.run() opt_parameters = result.coords

MLCost.get_error_estimate()

Source

def get_error_estimate(self, opt_parameters, log_l_variation=0.5):

Get confidence interval on ML estimated parameters.

Parameters:

opt_parametersarray of floats

The optimum parameters as obtained from the ML method.

log_l_variationfloat, optional

The variation in the log-likelihood upon changing the values of the parameters within the confidence interval: log_likelihood(interval[0]) <= log_likelihood(opt_parameters) - log_l_variation log_likelihood(interval[1]) <= log_likelihood(opt_parameters) - log_l_variation

Examples

To get an estimate on the confidence interval of the optimum parameters, do e.g.:

error = pot.get_error_estimate(result)