Misalignment between done and want to do
- Obviously there is a mismatch between what we want to
do and what we have already done
- How do we get from where we are, to where we want to be?
- One way is to do simulations (the way we've been doing it), but
that is quite undirected
- Another way is to look at neural data, though this is hard to
interpret (incomplete, different granularity, etc.)
- We've tried to look at the mathematics (like Beurle) but have found
that the system is really complex, and the dual dynamics (short and
long term) is really hard.
- Simpler models provide insight into how the more complex CA
models function
- For this abstract we proposed Hopfield and SOM models.
- We've done a bit of work on Hopfield models and CAs since the
abstract was submitted.
- For instance, we've calculated networks that have highly overlapping
CAs.
- This calculation was started with a similar Hopfield calculation,
and gradually spread to a CA network.
- The hope is that this calculation can eventually be turned into
a system that learns these patterns from environmental stimulus.
