By the way, another reason blogging has been so light is because, well, I'm in graduate school now, and I'm kept pretty damned busy.

We're supposed to complete a first-year research project, and I'm specifically looking at bilateral symmetry in artificial neural systems.

Most researchers that work with ANNs just build them from scratch. I use an algorithm that evolves the connection weights and architectures, so they can grow more complex. Most encoding schemes that do this are direct encoding, which means that one artificial gene encodes for one piece of the artificial neural network. This isn't very efficient.

So I'm looking at an encoding scheme in which certain genes are duplicated as a mirror image (unfolded along an axis of symmetry), while still allowing asymmetrical parts of the network to exist.

The domain is symmetry detection of visual stimuli (in this case, capital letters...A, B, C). The symmetrical inputs model the inputs to the right and left visual fields. Basically, if the fixation point for your vision is on the midline of a symmetrical figure, all that needs to be computed is that the input is balanced to both halves of the network. That should be easier if the network itself is symmetrical.

That's the hypothesis, at least.

I'm also looking at the relative impact of inter vs. intra-hemispheric connectivity. Is it more important for the two hemispheres of a neural system to be highly connected within themselves, or to have high connectivity between the hemispheres? Or a balance?

Anyway, this is the kind of stuff I'm spending most of my time working on and thinking about...so if blogging comes in erratic spurts, you'll know why.