3. The `ni.model.pointprocess` Module¶

class ni.model.pointprocess.PointProcess(dimensionality)[source]¶

A Point Process container.

Usually generated by loading from a file or via ni.model.pointprocess.createPoisson()

addSpike(t)[source]¶: adds a spike to the point process, if it falls in the allowed range.

getCounts()[source]¶: Gives a (in most cases binary) time series of the point process.

getProbability(t_from, t_to)[source]¶: Undocumented

plot(y=0, marker='|')[source]¶

Plots the pointprocess as points at line y.

marker determines the color and shape of the marker. Default is a vertical line ‘|’

plotGaussed(width)[source]¶: Plots the pointprocess as a smoothed time series

ni.model.pointprocess.PointProcessFromSpikeTimes(times)[source]¶

class ni.model.pointprocess.SimpleFiringRateModel[source]¶

Uses just the firing rate as a predictor

compare(Data, Prediction)[source]¶

fit(data)[source]¶

loglikelihood(Data, Prediction)[source]¶

predict(Data)[source]¶

ni.model.pointprocess.createPoisson(p, l)[source]¶

This generates a spike sequence of length l according to either a fixed firing rate p, or a repeated sequence of firing rates if type(p) == np.ndarray.

It creates a ni.model.pointprocess.PointProcess

Example 1:

p1 = ni.model.pointprocess.createPoisson(0.1,1000)
p1.plotGaussed(20)
plot(p1.frate)

p2 = ni.model.pointprocess.createPoisson(sin(numpy.array(range(0,200))*0.01)*0.5- 0.2,1000)
p2.plot()
p2.plotGaussed(10)

p2.plotGaussed(20)
plot(p2.frate)

Example with multiple channels:

frate = (numpy.array(range(0,200))*0.001)*0.2+0.01
channels = 9

dists = [ni.model.pointprocess.createPoisson(frate,1000) for i in range(0,channels)]
#for i in range(0,9): dists[i].plotGaussed(10)
import itertools
spks = np.array([dists[i].getCounts() for i in range(0,channels) for j in range(0,99) ])
imshow(-1*spks)
set_cmap('gray')

Will generate:

(A plot of spikes)

ni.model.pointprocess.plotGaussed(np.array([dists[i].getCounts() for i in range(0,channels)]).mean(axis=0),20)
plot(dists[0].frate)

ni.model.pointprocess.getBinary(spikes, min_length=1)[source]¶: Gives back a binary array from an array of spike times. The maximum for each bin is 1.

ni.model.pointprocess.getCounts(spikes)[source]¶: Gives back an array of spike counts from an array of spike times. If the output is suppsed to be a Binomial, use getBinary instead.

ni.model.pointprocess.interspike_interval(spikes_a, spikes_b=False)[source]¶

ni.model.pointprocess.plotGaussed(data, width)[source]¶

p2 = ni.model.pointprocess.createPoisson(sin(numpy.array(range(0,200))*0.01)*0.5- 0.2,1000)
p2.plot()
p2.plotGaussed(10)

ni.model.pointprocess.plotMultiSpikes(spikes)[source]¶

spikes is a binary 2d matrix

Generates something like:

ni.model.pointprocess.reverse_correlation(spikes_a, spikes_b=False)[source]¶

ni 0.2 documentation