Growing Axons Evolving L-systems

A. Carrascal, D. Manrique, D. Pérez, J. Ríos, and C. Rossi (Spain)

Keywords

Genetic algorithms, neural networks, medicine applications, LSystems.

Abstract

We propose a new technique for building neural networks which takes inspiration from the biology of neural cells. In our model neurons are fixed in space, and their connections are grown according to a given rule, for building any type of network without training. The growing rule is encoded in each cell's genetic code, and the strength of the connections depends on their length. Rules are generated using an evolutionary process. This process mimics the brains innate capabilities codifying this complex biological process with very few elements. Our first experimental results, both on laboratory tests and on a real world case, confirm the effectiveness of the proposed approach.

Important Links:



Go Back