Loading ...
Sorry, an error occurred while loading the content.

6483CFP on Neuroevolution of Indirect Representations of Neural Networks

Expand Messages
  • Julian Miller
    Jul 2, 2014
    • 0 Attachment
      This special issue of Natural Computing seeks high quality original work relating to the evolution of indirect representations of
      artificial neural networks (ANNS). Papers submitted can be original research papers or survey papers.

      In nature biological neural networks are created through an exceptionally indirect process. A single set of evolved genes in a single fertilized ovuum produces a network with an enormous number of cells and connections. In other words, the genotype to phenotype mapping is extremely indirect and complex. In addition, the process is ongoing during the lifetime of the individual. 

      Indirect genotype representations of ANNS do not use a one-to-one mapping of weights and connections to genes
      in the genotype. Typically, such mappings may be generative or developmental in nature, highly redundant or compressed. Usually, when such mappings are evolved small genetic changes can induce large scale changes in the networks (phenotypes). For instance a single gene change may cause multiple weights to change, neurons being disconnected and others connected, or large scale topological changes in the network.  The aim of such mappings is to make ANNs more evolvable and often to scale up to large networks capable of solving complex and large scale problems. Indirect mappings may be used in many kinds of neural networks: non-spiking, spiking, recurrent, etc.

      Relevant topics for the special issue include:

      Generative or developmental neural networks;
      Highly compressed genotype-phenotype mappings of neural networks;
      Highly redundant mappings in which many genes may be ignored in building the neural network;
      Evolutionary representations of neural networks that include greater biological plausibility;
      Evolutionary representations which include an encoding of neural learning rules;


      To submit a paper follow the instructions at:


      Please make clear that submissions are intended for the special issue entitled:

      "Neuroevolution of Indirect Representations of Neural Networks"

      Submission system opened: June 1, 2014.

      Closing date for submissions: September 1st, 2014.

      Informal inquiries can be emailed to  the guest editor:

      Dr. Julian Miller
      Department of Electronics
      University of York