Neural Model on Chip

The first task is to put the neural model and its associated learning rules onto PyNN and then the chips.
It's scheduled to take 12 months from April 2014.
We've already got a LIF version of it working on PyNN.
The first step is to put the CABot3 FLIF model onto PyNN; that's due by the 1st of June.
This might be a problem because fatigue is not a standard mechanism.
Hopefully, we can take one of the existing PyNN neural models and set its parameters to work.
A second problem is standard long-term learning. This needs to be added, but should be straight forward.
Short-term potentiation is also needed for binding in parsing.
There is some PyNN support for that, so we're optimistic that the Oct. 1st milestone will be met.
The chips should be available by then, so we'll try to put the models on the chips.
We also need the second fatigue variant.
We can build our own PyNN backend (like NEST) but the chips won't necessarily take that as input.
SpiNNaker we can manage as we've spoken with Furber about it for years. HICANN is more speculative.
All the primitives need to be on the chips by April 1st 2015.