da42c137d164ab56a1f1919c8f0ee21e6172f120,cde/evaluation/GoodnessOfFit.py,GoodnessOfFit,hellinger_distance_monte_carlo,#GoodnessOfFit#Any#Any#,134

Before Change

    

    if y is None:
      y = np.random.random(size=(n_samples, self.estimator.ndim_y))
    assert y.ndim == 2 and y.shape[1] == self.estimator.ndim_y

    P = self.probabilistic_model.pdf
    Q = self.estimator.predict

After Change

    
    assert x.ndim == 2 and x.shape[1] == self.estimator.ndim_x

    distances = np.zeros(x.shape[0])

    for i in range(x.shape[0]): // iterate over x values to condition on
      P = self.probabilistic_model.pdf
      Q = self.estimator.predict

      samples = stats.cauchy.rvs(loc=0, scale=2, size=(n_samples, self.estimator.ndim_x))
      f = _multidim_cauchy_pdf(samples, loc=0, scale=2)

      p = np.sqrt(P(x[i,:], samples))
      q = np.sqrt(Q(x[i,:], samples))

      r = (p - q)**2

      distances[i] = np.sqrt(np.mean(r / f)/2)

    assert distances.ndim == 1 and distances.shape[0] == x.shape[0]
    return distances

Instances