rules.py 文件源码

def flat_prior(self):
        """ Evaluate log-probability of each primitive in the population.

        Return value is properly normalized.

        In this base class, we implement a version where the 
        distribution over primitives is static; subclasses will 
        reevaluate this at each call based on the values of variables and 
        parameters.

        """

        raw_weights = np.zeros(len(self.rule_population))
        normalization = logsumexp(raw_weights)
        return raw_weights - normalization
###