The Goloko Neural Network for Vision - Analysis of Behavior

In this paper, some experiments with the GOLOKO neural network for vision are presented. The simulation of the GOLOKO was allowed by a new neural simulator called SiMoNNe which was developed in purpose of this simulation. The SiMoNNe simulator provides many useful features such as an easy and powerful development and debugging of complex neural network structures. The experiments described in this paper demonstrate that the GOLOKO neural network containing backward and associative connections converges faster than a network with feed-forward connections only. keywords: neural networks, computer vision, CALMSOM, GOLOKO