Results on Theory

There is not a well founded theory on how CAs work, though attractor nets are a good start.
We're hoping to understand the theory of CAs to help us solve our current problems and eventually go on to solve a host of problems
One area that seems wide open is the learning rule. We're wed to Hebbian learning, but there are a range of possible rules. We've done work on compensatory learning which leads to a range of good behavior like CAs spreading into unstimulated areas, and improved attractor dynamics.
We also have a proof that the capacity is greater than the standard proof for attractor nets. We're working on a manuscript now.
We've recently solved a problem of CA fractionation that we call the Mule Problem. We've done this by changing our neural fatigue model. We've submitted it to NCPW but will probably do a better experiment and submit it to Neural Computation.