Improvements
- I'd like to integrate the triplets into the system so that it's
learned automatically. I think that's the big win.
- I'd also like to have a second system that learned rules. We've
done stuff like that using reinforcement learning, still in neurons.
- This would give us a lot of scope for exploring
models of categorisation.
- Finally, I'd like to put it on Nest and SpiNNaker. This stuff
all runs on my CANT simulator. It should be relatively straight
forward to put it on the HBP systems.
- At least with SpiNNaker, this should enable the system to go
a lot faster.
