2022 11 07 Open NEST Developer Video Conference

previous | list | next

Agenda

1. Welcome
   • Simulate with EBRAINS (ongoing)
   • NEST Hackathon
   • HBP CodeJam #13
2. Review of NEST User Mailing List
3. Project team round
4. In-depth discussion
   • Eligibility Propagation (E-PROP) in NEST

Mailing list

• https://nest-simulator.org/mailinglist

Project team round

Here we discuss topics that need broader attention, for example questions that came up but are outside a single project's scope, larger planned changes/PRs that affect all teams or pending work that is blocked by external factors.

1. Models / NESTML
   • nothing to report, work as usual
   • tutorial in the "Simulate with EBRAINS" workshop
2. NEST GPU
   • first working version of cmake based installation
   • aligning Python interface to NEST
3. PyNEST
   • no report
4. NEST Desktop
   • no report
5. Kernel
   • ongoing revision of event delivery mechanisms
   • found possible performance improvement for connection exchange, to be investigated further
6. Installation
   • nothing to report
   • reviewing PR in nest-gpu
7. Infrastructure
   • Spack installs on Fenix sites now have a spack test nest available
8. Documentation
   • nothing to report
9. EBRAINS
   • Satellite events at HBP Summit 2023 (March)
     • deadline for proposals 21 Nov

In-depth discussion

Eligibility Propagation (E-PROP) in NEST

(Presentation by @akorgor and [@jespinoza])

Discussion

Is it possible to prevent delivery of (specific) secondary events?

• currently update once per time-step is quite wasteful
  • looked into delivery of secondary events
• communication code could maybe provide suitable mechanisms for different time interval communications

Is it possible to declare "shared" variables that can be used by neurons and synapses?

• similar problems in the Urbanczik-Senn model where certain values need to be equal in neurons and synapses
• maybe implementation of "shared" variables could be done in the kernel (?)
• NESTML could make sure that implementation is always consistent
  • depends on the nature of the variables, especially if the implementation is in the neuron or synapse properties
  • could also be a kernel property (get/set kernel status), similar to Structural Plasticity Manager, set in dictionary from one place, read out on the others.
• currently much of the users interaction is on the level of front-end dictionaries (Python)

Is it a problem that the only supported time resolution is 1 ms?

• …

~ discussion continued at the upcoming Hackaton ~