Rule Based Systems in Neurons
- I had an old student who was interested in doing some neural stuff.
I had thought that you could implement a rule based system in neurons,
and that it, along with an associative memory, would be a good initial
system for a neural cognitive architecture.
- He was interested, so I got him going with the neural finite state
automata stuff I'd been doing.
- He wrote a (python) system that converts a rule-base and facts to
a neural topology. The initial facts are turned on and the
rule based system runs in neurons.
- It works in NEST, but not SpiNNaker yet.
- We're checking that it works, and it still has usability problems,
but I think its Turing complete (given enough neurons).
- We're putting a paper together for SGAI (Cambridge in December).
- I've got an EPSRC neuro-cognitive architecture proposal in mind,
but that's not going to be very soon.
