Self-organizing ontogenetic development for autonomous adaptive systems - a dynamic perspective
Róbert Kozma, Derek Harter, Walter J. Freeman, Stan Franklin · 2002
The capacities of animals for pattern classification using biological neural mechanisms far surpass any existing artificial devices - advanced devices for pattern recognition using neural networks and statistical algorithms implemented with digital computers make use of certain biologically motivated principles of information processing. A major impediment to further development of biologically inspired computing tools is the lack of theory on how large masses of neurons interact to produce an emergent collective behavior. We are interested in developing models of ontogenetic development that capture some of the flexibility and power of biological development. In this paper we present a testbed for the creation and testing of models of development that we have created. We present some results on standard neural networks in learning to perform this task and discuss future plans for developmental models in this environment.