2016 09 19 Open NEST Developer Video Conference

previous | list | next

1. Welcome

2. Organizational

   1. Github: New tools for code reviews Start using the review tools, helps to open up the write permissions for the branches,
   2. More structured PR and Issue reviews prepare a foccussed list with special PR/issue highlighting (-> @terhorstd)
   3. Subnet removal code sprint (see #480) discuss date when to do the sprint with relevant people, not do this at the HBP-Summit.

3. Design Decision: How to handle input for conductance-based models with multiple synapses (see #261, #439 and #477)

   • sign is used for selecting the right input buffer, conductance based still possible, multi-synapse models in conductance based models can be modeled as two different ports (excitatory/inhibitory)
   • problem in multi-synapse models
     • multiple ports, and for each port two buffer
     • better: one port with reversal-potential
     • note: no issues implementing this in NESTML
     • regarding #477, not clear what the proper solution would be
     • precise models need spikes to arrive in the same buffer. do we need sth. coherent between all models? leave precise models separated for the time being; maybe revise later. otherwise multiple buffers would need to be searched => decision: one buffer for each port in multiple-synapse models
     • Decision:
       • For all multisynapse models, there will be exactly one spike input buffer per receptor port.
       • For current-based multisynapse models
         • each receptor port has its own synaptic time constant(s)
         • input spikes with negative weights will be added to the input buffers with their negative weight, representing hyperpolarizing currents
       • For conductance-based multisynapse models
         • each receptor port has its own synaptic time constants(s) and a single reversal potential
         • input spikes must have strictly non-negative weights
     • Implementation: Project 1

4. Philip

   • new device "weight recorder" for synapses
   • normal recorder can not get the post-synaptic ID.
   • changed event class to have a new field
   • Q: do we want the WR in the common synapse properties?
   • Q: every event has another GID field: 'receiver gid', ok? -- better not put it into the base class.
   • good solution until NEST-io, do PR now, and change later

5. Some ideas for representing parameters, see #488

   • syntax examples, exact syntax needs to be discussed -> setup a work-group to discuss this on the NEST user workshop (Responsible: @heplesser)

6. AOB: tomorrow new MAC OSX, someone want's to do testing?

7. Review of NEST User Mailing List (Skipped due to lack of time)

8. Review of open Github Issues (Skipped due to lack of time)

9. Review of open Github Pull Request (Skipped due to lack of time)