Future Work
- Figure out competition:
- Diehl and Cook do it with a dynamic threshold.
- Long refractory periods might work.
- I think we haven't quite got balanced firing with
inhibition right.
- What should happen is that the receptive field
of each output neuron moves to a different part of the
presented input space like a SOM.
- Also need to figure out how the firing times interact
with this.
- Combine with work on recurrence and Cell Assemblies. We've done
a lot of category learning with spiking nets but at a 10ms time
step with compensatory learning.
- The STDP timing should interact with this, hopefully in a positive
way.
- Perhaps look at spiking neurons for long short term memory.
- Ultimately, we'd like to know who the brain does it.
