This mimics variance from Lowpass filtered spikes going into
an integrative membrane (e.g. SpikingRectifiedLinear).
TODO: The variance of this noise model is higher than with AlphaRCNoise and LowpassRCNoise. I'm not sure if this is a bug; in fact, I actually think that the variance for those two models is too low. I need to revisit this and think more about how these noise models should work, specifically how the postsynaptic membrane should apply. One thing to compare with would be the noise models without the membrane included at all (i.e. just filtering the spikes). Also, rather than applying the neuron membrane over long periods to the filtered spike train, we could look at applying it over short periods only, perhaps the period of the maximum firing rate we expect in the network.
This mimics variance from
Lowpassfiltered spikes going into an integrative membrane (e.g.SpikingRectifiedLinear).TODO: The variance of this noise model is higher than with
AlphaRCNoiseandLowpassRCNoise. I'm not sure if this is a bug; in fact, I actually think that the variance for those two models is too low. I need to revisit this and think more about how these noise models should work, specifically how the postsynaptic membrane should apply. One thing to compare with would be the noise models without the membrane included at all (i.e. just filtering the spikes). Also, rather than applying the neuron membrane over long periods to the filtered spike train, we could look at applying it over short periods only, perhaps the period of the maximum firing rate we expect in the network.