Fig. 15
From: Responses of Leaky Integrate-and-Fire Neurons to a Plurality of Stimuli in Their Receptive Fields
Model selection accuracy. Successful selection is defined as a DIC difference greater than 2, and the proportion of correctly identified models is calculated over 100 repetitions. Note that a not correctly identified model in most cases means that the DIC difference was smaller than 2, not that the wrong model was selected. Top left: proportion of correctly identified models with weights from 0.1 to 0.5 and perturbation size from 1 to 10 for RA data, using 10 spike trains. Top right: the same for PM data. Middle left: proportion of correctly identified models for number of spike trains of 6 to 18 and stimulus scaling from 0.6 to 1.4 for RA data, using a weight of 0.3 and a perturbation size of 3, shown in green in the top panels. Middle right: the same for PM data. Bottom left: the two stimuli curves (black and red) with perturbation size 3 (sinusoidal curve shown in blue) used for the cases shown in green in top panels. Bottom right: example spike trains following either RA or PM, using the two stimuli shown in the left, with weight 0.3