Abstract: There is predicated a decision making problem with the finite set of neural network training algorithms as alternatives against a set of that neural network parameters, being controlled and estimated. The rules for making decisions on selecting the training algorithms optimally are formulated.

Keywords: neural network, training algorithm, decision making problem, point-evaluation, interval-evaluation, Bayes — Laplace rule, optimal strategy.

ACM Classification Keywords: H.4.2 Types of Systems — Decision support, I.2.6 Learning — Connectionism and neural nets.

Link:

DECISIONS ON SELECTING THE TRAINING ALGORITHM OF THE NEURAL NETWORK WITH A SET OF ITS BEING CONTROLLED PARAMETERS

Vadim Romanuke

http://foibg.com/ibs_isc/ibs-28/ibs-28-p18.pdf