Abstract: The problem state and multicriteria optimization procedure of neural network classifier’s architecture is considered. The scalar convolution of criteria with nonlinear trade-off scheme is offered as a goal function. The search methods of optimization with discrete arguments are used. The neural network classifier of texts as an example is given.

Keywords: multicriteria optimization, neural nets, classifier.

ACM Classification Keywords: H.1 Models and Principles – H.1.1 – Systems and Information Theory; H.4.2 – Types of Systems; C.1.3 Other Architecture Styles – Neural nets

Link:

THEORETIC-EXPERIMENTAL MULTICRITERIA METHOD FOR NEURAL NETWORK CLASSIFIERS’ ARCHITECTURE

Albert Voronin, Yuriy Ziatdinov, Anna Antonyuk

http://www.foibg.com/ibs_isc/ibs-15/ibs-15-p06.pdf