Publikace detail

Piecewise-Linear Neural Networks Training Using Gradient Descent Approach

Autoři: Gago Lumír | Doležel Petr

Rok: 2016

Druh publikace: ostatní - článek ve sborníku

Název zdroje: Automatizácia a riadenie v teórii a praxi 2016

Název nakladatele: Technická univerzita v Košiciach

Místo vydání: Košice

Strana od-do: 1-7

Tituly:

Jazyk	Název	Abstrakt	Klíčová slova
eng	Piecewise-Linear Neural Networks Training Using Gradient Descent Approach	This article focuses on comparison of learning of artificial neural network with a hyperbolic tangent activation function and linear saturated activation function in hidden layers by backpropagation algorithm with momentum. For evaluating of learning characteristics, two experiments with specific value of iterations are performed. The results are analyzed using boxplot graphs for each pair of artificial neural networks. An empirical result discussed comprehensively at the end of the paper is, that the approximation qualities of both networks under examination are similar.	Piecewise-Linear Neural Networks; Backpropagation Algorithm

Search