Fig 7. Active parameter selection with Gaussian processes.

Data: ROI from a scan field with bipolar cell terminals in a retina expressing iGluSnFR, recorded using a spiral scan configuration. Model: Product of RBF kernel (time) with composite RBF kernels (frequency and contrast), 500 inducing inputs, 50 iterations per fit, best of 4 fits per model. a: Control stimulus consisting of 90 parameter sets of frequency and contrast. Uncertainty in each region computed as the sum of the latent uncertainty for a GP estimated under all parameter configurations. The chirp response for this ROI is shown above. The completed GP model for the sine response is estimated over the full dataset, the model inference for the sinusoidal chirp components is shown. b: Active parameter selection using GP with corresponding inferences for chirp stimulus.