Software WG taskforce: simulator selection resource
Meeting 5: Wed 25 Jan, 1400 UTC
Agenda
Attendees
- Brent Huisman
- Marcel Stimberg
- Dinara Issagaliyeva
- Ankur Sinha
- Reema
- Eliane Rodrigues
Next meeting
Next meeting: Wed 8 Feb, 1400 UTC
Frontend ideas:
- Have glossary, but use it in reverse
- eg. use "1k-1M cells" and have glossary/tooltip saying that this can be refered to as mesoscale
- ranged slider for network scale
- sort results based on nearness to range rather than simple ascending
Prototyping:
- come up with simple designs just to get a layout of how we're going to expose the functions
- example tools
Implementation ideas:
- Example of Python script to generate an interactive but static website from json data
- example: brian uses brian tools for things, and arbor-GUI uses arbor
- have "defined relationships": when arbor is listed in the result, it shows "related tools: Arbor-GUI (GUI)", and when Arbor-GUI is in the result, it shows "related tools: Arbor (simulation engine)"
Questions/notes while tagging
Ankur
- network scale will need to be defined in our glossary
- artificial neuron
- single compartment neuron
- sub cellular
- single cell
- small scale networks. 1-1k cells. 1e0-1e3
- meso scale networks. 1k-1M cells. 1e3-1e6.
- large scale networks. 1M-1B+ cells. 1e6-1e9+
- use simpler terminology (log scale): one, hundreds, thousands, millions and more?
- then no slider ;_;
- prefer multiple choice?
- Since simulators usually scale down, we can simplify to "up to"
- The "scale" could be numerical and the jargon of "small, meso, large, etc" could be prsented as an "info" pop up(?) - this means that simulators DO need to be annotated with scale terminology
- file formats:
- have tool tip to indicate to users that these are formats that are "natively" supported by the tool
- in: do we mean what data files it can take in? A little hard to list if it's using something like Python as an interface language because it can ready pretty much anything
- [BH] Examples: model descriptions in the morphological world can include
- Morphology formats (swc, asc, NeuroML are some popular ones).
- Mechanisms (ie. nmodl, NeuroML) Writing compilers for this is no easy feat :)
- Simulation 'formats' like NeuroML, Sonata. May contain the above plus more (networks, in HDF5 for instance, parameter data, sometimes also experimental outputs)
- same for output: most tools can output plaintext, but with libraries, they can output in any format pretty much
- [BH] Some projects list eg NWB as 'built-in' output format. You're right, you can always do that yourself.
- Dependencies between projects
- link associated tools somehow
- Another tag useful for users: analysis features?
- what built in analyis features does the tool have?
- visualisation/plotting of morphologies, spike trains, rates, concentrations etc.
TODO
- complete data
- free form tags
- fixed/required tags
- fileformat
- model complexity
- model size
- file format
- related/associated software
- have a sketch/proof-of-concept
- think about technology