Feedforward Neural Network Methodology
Springer, New York, 1999
by
Terrence L. Fine
Emmanuel J. Candes, Statistics, Stanford University, in Jour. American Statistical Association, June 2000, p. 682.
Contributor: Fine
Due to a Latex2e error in the manuscript, pages 142 and 143 are shortened, although no material is omitted.
Contributor: Fine
Eq. 5.6.1 on p.173 for the quasi-Newton BFGS update of the approximate inverse Hessian is perhaps clearer if we re-index by replacing the left-hand side M_{k+1} by M_k, the right-hand side M_k by M_{k-1}, and q_k =g_k -g_{k-1}, p_k =w_k -w_{k-1}.
See the pdf document A note on architecture selection.
Contributor: Peter Bartlett
Definition 4.3.2 on p.92 should be corrected by replacing
\exists f \in G by \exists f \in F
Date: 5/27/00
Contributor: Peter Bartlett
p158, (wk - w^*) in the displayed equation at the bottom should right multiply the matrix
Last revised: 6/16/00