Dimensionality reduction methods in Nengo


Has anyone had experience with processing the spike data with some dimensionality reduction techniques (i.e., PCA, GFPA, etc)? How would you go about extracting the spike data of an ensemble to use it for such a technique?


Hello! You can use a Probe to probe ens.neurons; there’s an example here that shows how to do this and use the rasterplot helper function Many neurons — Nengo 3.0.0 docs (the A_spikes probe)

1 Like

Okay thanks! Is there a way to get instantaneous firing frequencies from the spike train data? Or a built in function or something to get them?


Best way is to just analyze the spike trains. You’ll have 0s and 1/dt values in the spike train, so to calculate the average firing rate per second can just sim.data[probe_spikes] * dt / length_of_sim (this is assuming that you haven’t set up the neurons to have a custom spike amplitude)

1 Like