Measuring GNG Topology Preservation in Computer Vision Applications

Self-organizing neural networks try to preserve the topology of an input space by means of their competitive learning. This capacity has been used, among others, for the representation of objects and their motion. In addition, these applications usually have real-time constraints. In this work we have study a kind of self-organizing network, the Growing Neural Gas with different parameters, to represent different objects. In some cases, topology preservation is lost and, therefore, the quality of the representation. So, we have made a study to quantify topology preservation to establish the most suitable learning parameters, depending on the kind of objects to represent and the size of the network.