Undesired behaviour recording spikes

[Melissa]

Hello everyone,

I'm trying to run a simple Brunel network on SpiNNaker and calculate the average firing rate in Hz. However, I've encountered some weird behaviour:

When using IF_curr_alpha Neurons are used, the network does not seem to be active at all no matter which weights I use. At least no spikes recorded.

When using IF_curr_exp neurons, another problem occurs: Only if the weights are very low, the spikes seem to be recorded correctly since, if I calculate the average firing weight using 1000*(pop.mean_spike_count()/simtime) with the weights being a little higher, and I also create a plot with the spikes, the two don't match up. 

For example:

In the last run over 2300ms, the average firing rate of around 23.7 was calculated. Viewing the plot, it is pretty obvious that those are on average way more than 54.51 spikes per neuron. 

Has anyone encountered a similar problem and/or knows how to handle this and why this problem occurs? 

Thanks in advance!

Best wishes,

Melissa :)

[Andrew Rowley]

Hi,

Apologies for the delay in replying. Can you send me some example scripts where you are experiencing these issues? I can then see if any of it is expected for some reason, or if there is a bug somewhere.

Thanks,

Andrew :)

________________________________________
From: spinnak...@googlegroups.com <spinnak...@googlegroups.com> on behalf of Melissa <melissa...@live.com>
Sent: 28 March 2025 08:47
To: SpiNNaker Users Group
Subject: [SpiNNaker Mailing List] Undesired behaviour recording spikes

--
You received this message because you are subscribed to the Google Groups "SpiNNaker Users Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spinnakeruser...@googlegroups.com<mailto:spinnakeruser...@googlegroups.com>.
To view this discussion, visit https://groups.google.com/d/msgid/spinnakerusers/29b7d122-3591-4604-aa0c-70278de0b560n%40googlegroups.com [groups.google.com]<https://urldefense.com/v3/__https://groups.google.com/d/msgid/spinnakerusers/29b7d122-3591-4604-aa0c-70278de0b560n*40googlegroups.com?utm_medium=email&utm_source=footer__;JQ!!PDiH4ENfjr2_Jw!AYczR_Rs0UuUwxbUcr-XIint8MoDqdbJlxD_w-ssQPSLgDId8-1NYv83joVjleN3URTCtmdizsNHwB3oXn_-k4xNMOjILu4RrH4$>.

[Andrew Rowley]

Hi,

I just wanted to add that I tested some things using our own Balanced Random network, which I think is a Brunel-style network. This example is here:
https://github.com/SpiNNakerManchester/PyNNExamples/blob/master/balanced_random/balanced_random.py

With that network, I added a "mean count" to the script after reading back the spikes:
mean_count = pop_exc.mean_spike_count()

Then just before drawing the graph, I added a check:
n_spikes = sum(len(s) for s in data.segments[0].spiketrains)
mean_spike_count = n_spikes / pop_exc.size
print(mean_count, mean_spike_count)

These seemed to work out correctly. This continued even if I increased weight_exc.

I also then tried replacing IF_curr_exp with IF_curr_alpha. This only worked once weight_exc was quite high e.g. a weight of 0.5 wasn't enough to see any activity really, but 0.6 started to show activity, and 0.7 made that activity stronger.

Andrew :)

________________________________________
From: Andrew Rowley <Andrew...@manchester.ac.uk>
Sent: 02 April 2025 15:51
To: Melissa; SpiNNaker Users Group
Subject: Re: [SpiNNaker Mailing List] Undesired behaviour recording spikes