Thesis presented for the Licentiate degree:

Motion Detection by Cooperation

A case study of the time dimension in neural networks

Robert Pallbo © 1994

LU-CS-TR:94-125 LUNFD6/(NFCS-3086)/1-71/(1994)

Robert Pallbo » Academic » Papers » Licentiate thesis

Contents	Abstract [below] Introduction [full text in pdf] Visual motion detection based on a cooperative neural network architecture Motion Detection - A neural model and its implementation Ontogenesis in neural networks

Abstract	This is a thesis for a Swedish licentiate degree. It consists of three papers along with a general introduction. The work described is the construction of a model of motion direction detection by the use of a neural network. The model is unusual in that it does not make use of any intermediate units. Instead, the input units are directly connected to the units that constitutes the detectors. The capacity of the network to detect motion is an emergent phenomenon which is dependent on the presence of spontaneous activity among the units. In addition to a presentation of the model and some results from simulations, it is discussed how the methodology can be used in applications other than motion detection.

Copyright © 1995-2005, Robert Pallbo