GROWING NEURAL NETWORKS

Growing neural  networks  are  networks  which  are constructed by executing 
genetic instructions contained in a genotype. These  instructions  and their 
products interacts non-linearly to eventually determine the mature structure. 
We present simulations in which the mapping  from  genotype  to phenotype is 
instantaneous and simulations in which it develops in time during  a segment 
of  an  individual's lifetime,  i.e.  the  individual's  developmental  age, 
allowing  us  to  study  both  neural evolution  and neural development. The
results  shed  some  light  on (a) why  modular architectures  are likely to 
emerge, (b) why  similar successions  of  stages  tend  to  appear  in  both 
evolution  and  development,  and (c)  why  a developmental age is preserved 
evolutionarily although  the mature state  may  appear to  be more efficient 
from the point of view of fitness.