ann_FF_VHess — multiplication between a "vector" V and Hessian
VH The result of multiplication - hypermatrix with the same layout as W. x Matrix of input patterns, one pattern per column. t Matrix of target patterns, one pattern per column. N Row vector describing the number of neurons per layer. N(1) is the size of input pattern vector, N(size(N,'c')) is the size of output pattern vector (and also target). W The weight hypermatrix. V The "vector" by which Hessian have to be multiplied, actually is a hypermatrix with same layout as W (is from same space). dW Size of "finite difference". af The activation function to be used. This parameter is optional, default value "ann_log_activ", i.e. the logistic activation function. err_deriv_y the name of error function derivative with respect to network outputs. This parameter is optional, default value is "ann_d_sum_of_sqr", i.e. the derivative of sum-of-squares.
This function calculates the product between a vector and Hessian trough a (fast) finite difference procedure. The error gradient is calculated (trough backpropagation) at W+dW*V and at W-dW*V and then VH is calculated directly from here (this require only two backpropagations, explicit computation of Hessian is avoided).