Schematic description of the training of the network to obtain the covariance matrix and the derivative of the mean vector that are used in Eq. (21) to obtain the Fisher information matrix. The top, middle, and bottom lists of images form what we call the training set and are generated with fiducial parameters θ_fid (with θ_fid + Δθ_α and θ_fid − Δθ_α, for each α ∈ {1, …, p}). From the first collection of images, we compute the covariance matrix, and from the second and third collections of images, we compute the derivative of the mean vector for each α ∈ {1, …, p}.