Artificial neural networks (ANN) are well known and widely used from decades. One of the most popular is the multilayer perceptron (MLP). MPL’s general characteristic is that it has more than one layer. The most used topology of MLP has three layers (input, hidden and output). Layers are fully-connected, between each other (the input with the hidden layer and the hidden layer with the output). The general disadvantage of this topology is the lack of recurrent connections. The common usage of MLP is to solve only a single task. This research addresses these two common features of the MLP.
Ivan Blagoev
Janeta Sevova
Kolyu Kolev
Bulgarian Academy of Sciences, Sofia
Bulgaria
e-mail: kolev_kolyu@yahoo.com
Abstract:
Key words:
artificial neural networks
multilayer perceptron
topologies
Section:
Topics: