Fig. 6.

Simulation results comparing the posterior mean estimates of the STRF estimated using a white vs. optimized Gaussian process. The STRF used to generate the data was the STRF fitted to the neuron in Fig. 3(a). In this case the posterior mean estimates seem to converge to the true STRF at the same qualitative rate; however, as shown in Fig. 7 below, the uncertainty of our estimate shrinks more rapidly under the optimized Gaussian process design