Skip to main content
. 2011 Jan 11;6(1):e16104. doi: 10.1371/journal.pone.0016104

Figure 2. A GLM as a neural encoding model.

Figure 2

(A)–(B) Estimated parameters for an example auditory midbrain neuron. (A) STRF. (B) Exponentiated post-spike filter, which may be interpreted as a spike-induced gain adjustment of the neuron's firing rate. It produces a brief refractory period and gradual recovery (with a slight overshoot). (C) Estimate of the nonlinearity transforming linear input to instantaneous spike rate (black points), for the same example neuron (Chichilnisky 2001). The nonlinearity represents the probability of observing a spike for each value of net linear input (b+k*x+h*r). An exponential function (grey line), the assumed nonlinearity for the model, provides a reasonable approximation to this function. (D) Spectrogram (x) of one example song used in the experiments. (E) Stimulus filtered by STRF, k*x. (F) Recorded (gray) and predicted (red) raster plots in response to the validation stimulus shown in (D).