Using Marker-Based Genetic Encoding Of Neural Networks To Evolve Finite-State Behaviour (1991)
A new mechanism for genetic encoding of neural networks is proposed, which is loosely based on the marker structure of biological DNA. The mechanism allows all aspects of the network structure, including the number of nodes and their connectivity, to be evolved through genetic algorithms. The effectiveness of the encoding scheme is demonstrated in an object recognition task that requires artificial creatures (whose behaviour is driven by a neural network) to develop high-level finite-state exploration and discrimination strategies. The task requires solving the sensory-motor grounding problem, i.e. developing a functional understanding of the effects that a creature's movement has on its sensory input.
In Francisco J. Varela and Paul Bourgine, editors, Toward a Practice of Autonomous Systems: {P}roceedings of the First {E}uropean Conference on Artificial Life, 255-262, Cambridge, MA, 1991. MIT Press.

Brad Fullmer Undergraduate Alumni
Risto Miikkulainen Faculty risto [at] cs utexas edu