70 Computation of Gradient and Hessian in Feed-Forward Neural Networks: A Variational Approach
Download citation file:
- Ris (Zotero)
- Reference Manager
In this paper we derive explicit expressions for the gradient and the Hessian of the objective function (which in the sum of the squared errors) with respect to the weight matrices in a general feed-forward neural network using the standard methods for computing the first-order and second-order variations of a scalar-valued function of a vector.