The Permutation Problem for multi-layered networks.

The Permutation Problem and the effective optimising of Artificial Neural Networks have been investigated [SH] for fixed architecture single-layer feed-forward networks. It is expected that the theory presented will apply equally well to multi-layered networks, though it will need to be extended in order to precisely determine how multiple layers will affect the probability of the problem occurring. The research will enable connectionists [Neuroevolutionary] to comprehend (quantify) the Permutation Problem with respect to multi-layered networks – this work will be used to enhance connectionist [Neuroevolutionary] systems. The research aims to demonstrated and amplified through a set of experimental steps, enabling us to access the strength and weakness of the approach. The contribution the research will make to the connectionist [Neuroevolutionary] domain is the Permutation Problem for multi-layered networks is an important research issue and to demonstrate the basic feasibility of the investigation.**

Reference: [SH] Haflidason S., (2010), “On the significance of the permutation problem in neuroevolution,” PhD thesis: in the School of Computer Science, The University of Manchester.