Committee Methods In Recognition Neural Networks

2006
Aleksandrs Glazs, Maksims Alekseičevs, Tatjana Pudža

There are committee methods in recognizing neural networks research results shown in the work. It is shown that by network architecture complication voting algorithm gives better results than the average algorithm. Weighted voting and weighted average voting procedure application with the compulsive training cycle interruption does not provide a better recognition result.

Keywords
neironu tīkls, komiteju lēmums, ansambļa vidējā lieluma algoritms.

Glazs, A., Alekseičevs, M., Pudža, T. Committee Methods In Recognition Neural Networks. Technologies of Computer Control. Vol.27, 2006, pp.24-29. ISSN 1407-7493.

Publication language
Latvian (lv)