2016 10 17 Open NEST Developer Video Conference

previous | list | next

1. Welcome

2. Information

   1. Brief report from the Human Brain Project Summit
   2. NEST User Workshop

3. Review of NEST User Mailing List

   • NESTML workshop announced
   • contact with robotics from Munic -> discuss at NEST Workshop in Karlsruhe
   • calcium concentrations recorder
   • question for guidance on post synaptic inhibition, needs more information

4. Design Decision: defaults for multisynapse models + State_::set() (#510 and other tasks in multisynapse revision project)

   • negative weights make no sense for conductance based neurons
   • how many synapses should multi synapse models have per default? zero/one? Which default parameters?
     • since there are already defaults for single-synaptic neurons, keep it similar for multi-synaptic.
     • DECISION unless you specify something, multi-syn models should behave exactly as the "normal" ones.
   • Should the conductance be settible or leave this to the model to decide?
     • C: why should we make a difference for something, just offer setting for all state variables
     • C: once we write models with NESTML, this can not easily be chosen per model.
     • C: no significant cost for making the state variables settable. NESTML already has "supress" keyword, that already takes care of this -> non-issue
     • There should be reasons for hiding a state-variable, since generally user is interested in states.
     • DECISION: no need to explicitly make this consistent across models, will be handled when NESTML is available.

5. Design Decision: remove aeif_*_dynamics_DT0 (linked to the remark by @golosio in #514)

   • DECISION: leave it in and put into an if-then, so it doesn't take compute time except for this decision, if the function is not used, maybe remove it.

6. Review of open Github Pull Request

   • #464: @jougs is at it, help can only be build where python is available (-> no issue)
   • #507 (change of growth factor 1.5): postponed to discuss with author
   • Review(er)s required: #518, #515, #513, #491 -> reviewers suggested/added
   • Waiting for second 👍 : #497 -> suggested/added, #485 (skipped)

7. Review of open Github Issues

   • #506 Voltage-dependent STDP rules
     • @suku248 will also look into this
   • #521 : probably due to compressed representation (to save memory), we should take a look at this
   • #506 @flinz
     • principally not a problem, pre-synaptic spikes trigger the processing
     • potentially eats a lot of memory if there is a synapse that never receives a spike (probably not that much of memory)
     • signaling through the post-synaptic side would be difficult
     • time-driven updates (as done by @suku248) could limit the length of archive lists.
   • #457 Data races
     • seems to be found in serial parts.... someone who wants to dig deep, can/should have a look at this.
   • #484 Python bindings crashing on OSX
   • #487 SLI dup function can cause segmentation fault
   • #477, #470