gaussian_conv_regressor.py 文件源码

def log_likelihood_sym(self, x_var, y_var):
        normalized_xs_var = (x_var - self._x_mean_var) / self._x_std_var

        normalized_means_var, normalized_log_stds_var = \
            L.get_output([self._l_mean, self._l_log_std], {
                self._mean_network.input_layer: normalized_xs_var})

        means_var = normalized_means_var * self._y_std_var + self._y_mean_var
        log_stds_var = normalized_log_stds_var + TT.log(self._y_std_var)

        return self._dist.log_likelihood_sym(y_var, dict(mean=means_var, log_std=log_stds_var))