Have more cell types builtin (e.g. Izhikevich) to improve NeuroML compat

[brenthuisman]

NeuroML relies on the simulator to have implementations for cell kinds, one of which is izhikevichCell. The rather short list of point neuron models in Arbor has been mentioned before, so that makes two reasons to consider expanding the list. For reference: here (scroll down on the RHS) the list that NeuroML expects.

• [ ] Get feedback on a list of most wanted cell kinds.
  • [ ] Bring up in the upcoming Developer Meeting.
• [ ] Are there any cell kinds needed to show off our NML support? Anything that would to well in a publication?

Whoever reads this, please add your cell kinds in a comment.

[thorstenhater]

This is only a small step on the road to a larger project:

• allow users to bring their own Cell(Group)s #1605
  • this is either an extension to our current ABI or a new ABI aimed at cell groups
• unify probing across cell groups
  • started here for LIF #1263
• (optional and hard) extract cells from NML2 description and compile them to whatever #1605 comes up with.

[schmitts]

AdEx would be nice

[brenthuisman]

1605 is ambitious and long term. Earlier we advised people to write a mechanism to implement their own single compartment neuron, e.g. https://github.com/tetzlab/FIPPA