% Demo Code for reproduction
% Paper:Exact Stochastic Newton Method for Deep Learning: the feedforward networks case.
% Used dataset: MNIST

% Main code: ESN_mnist_demo.m
% functions list
    % feedforward: forward pass
    % backprop_sym: backward pass
    % grd_init: empty structure with the dimension of the gradient
    % grd_sym: cross correlation between the state and thee adjoint to compute the gradient
    % sigmoid: activation function Leaky ReLu
    % dsigmoid: derivative of activation function
    % d2sigmoid: second derivative of activation function (0) for Leaky ReLu
    % network_: initialize the networks structure using cells
    % Hsv: fast Hessian vector product.

% Instruction: Run mnist_demo.m, updates are displayed regularly,results are saved in .mat format in 'results' folder,
