hopsy.Model.compute_expected_fisher_information#

Model.compute_expected_fisher_information()#
deprecated:: 1.4

Use log_curvature() instead.

For some proposals, the expected fisher information will help converging faster as long as the gradient computation is not too slow. If you can not compute a useful or fast enough expected fisher information for your custom model, you can just return a zero matrix with the correct dimensionality (number of rows and cols each equal to number of parameters).

:param : :type : param x: Input vector :param : :type : type x: numpy.ndarray[float64[n,1]]

Returns:

  • return: The value of model.compute_expected_fisher_information(x)

  • rtype: numpy.ndarray[float64[n,n]]