A question to those of you who write ecf config scripts in your sleep:

is there an obvious way of writing the config script to include a local
learning step in each generation using a) a 'baldwinian' model; b) a
lamarckian model?

eg imagine the maxones case -  i just want to add, with probability of
application p, a steepest local ascent step without (in the case of a) and
with (in the case of b) the inheritance of accquired characteristics

also - is there a tie in with any common open source neural net packages
to allow for evolutionary training of neural nets?

thanks - and apologies for imposig on the good nature of the list again

regards
tony
-----------------------------------------------------------
SEE THE WICKED ROBOT INVASION MAP
http://www.wickedrobots.co.uk/
-----------------------------------------------------------
Mail tags:
-----------------------------------------------------------
Tony Hirst
mailto:[log in to unmask]
blog: http://micro-info.blogspot.com/

Dept. of ICT, Faculty of Technology
Open University, Walton Hall,
Milton Keynes, MK7 6AA, UK

Tel: +44 (0)19086 52789, m./SMS 07709 766223

http://robofesta.open.ac.uk/tony
http://www.robofesta-uk.org