Ring buffer sigma and spikes-per-second on a per synapse-type and per-population basis

[neworderofjamie]

Different populations have incoming spike trains with different statistics therefore it makes little sense to specify this globally (also the global parameters are copied into each vertex anyway!). Also, different synapse types have different statistics and separate ring-buffer shifts to scale them with, so ideally, you should also be able to specify ring buffer sigma and spikes-per-second separately for different synapse types.

[alan-stokes]

changes have ben done to spinnaker to allow changes of the spikes per second and ring buffer sigma from a top elvel pynn script.

[AlexRast]

Does this do it on a per-population basis? If so then I think there's a strong
case to be made as mentioned earlier regarding external I/O functionality to
migrate these 2 parameters into the subclass of vertices that actually use it
rather than making them global to all models.

On 19/08/15 15:23, Alan Stokes wrote:

changes have ben done to spinnaker to allow changes of the spikes per second and
ring buffer sigma from a top elvel pynn script.

—
Reply to this email directly or view it on GitHub
#132 (comment).

[rowleya]

Unfortunately, this has been added to take account of how PyNN 0.7 works (you pass the class and the initialization parameters, rather than an instantiated class). As I understand it, PyNN 0.8 passes an instantiated class, so at this point, we will have to change to a different interface for these parameters.

From: AlexRast [mailto:[email protected]]
Sent: 19 August 2015 17:14
To: SpiNNakerManchester/sPyNNaker [email protected]
Subject: Re: [sPyNNaker] Ring buffer sigma and spikes-per-second on a per synapse-type and per-population basis (#132)

Does this do it on a per-population basis? If so then I think there's a strong
case to be made as mentioned earlier regarding external I/O functionality to
migrate these 2 parameters into the subclass of vertices that actually use it
rather than making them global to all models.

On 19/08/15 15:23, Alan Stokes wrote:

changes have ben done to spinnaker to allow changes of the spikes per second and
ring buffer sigma from a top elvel pynn script.

—
Reply to this email directly or view it on GitHub
#132 (comment).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/132#issuecomment-132667966.

[AlexRast]

Isn't that the same thing (i.e. the initialisation parameters that are valid for
a population type are themselves dependent on the type of population being
instantiated)? So these can easily be asssociated with particular population
classes, which will be instantiated as the appropriate vertex subclass, while
other population classes neither take nor use these parameters.

On 20/08/15 10:18, Andrew Rowley wrote:

Unfortunately, this has been added to take account of how PyNN 0.7 works (you
pass the class and the initialization parameters, rather than an instantiated
class). As I understand it, PyNN 0.8 passes an instantiated class, so at this
point, we will have to change to a different interface for these parameters.

From: AlexRast [mailto:[email protected]]
Sent: 19 August 2015 17:14
To: SpiNNakerManchester/sPyNNaker [email protected]
Subject: Re: [sPyNNaker] Ring buffer sigma and spikes-per-second on a per
synapse-type and per-population basis (#132)

Does this do it on a per-population basis? If so then I think there's a strong
case to be made as mentioned earlier regarding external I/O functionality to
migrate these 2 parameters into the subclass of vertices that actually use it
rather than making them global to all models.

On 19/08/15 15:23, Alan Stokes wrote:

changes have ben done to spinnaker to allow changes of the spikes per second and
ring buffer sigma from a top elvel pynn script.

—
Reply to this email directly or view it on GitHub

#132 (comment).

—
Reply to this email directly or view it on
GitHubhttps://github.com//issues/132#issuecomment-132667966.

—
Reply to this email directly or view it on GitHub
#132 (comment).

[alan-stokes]

Different populations have incoming spike trains with different statistics therefore it makes little sense to specify this globally (also the global parameters are copied into each vertex anyway!).

This has been merged into master. the changing it for different synapse types, has not yet been completed.

[alan-stokes]

splitting the extra part into a new ticket. so we can record that 1/2 is done by oen release date and the other by dec.

[alan-stokes]

linked to #136