Fig. 5
From: Propagation of temporal and rate signals in cultured multilayer networks
Propagation of firing rate information. a Average PSTHs in layers 1–4 (left to right) compiled from trials sorted in five groups based on firing rates in layers 1 (see text and Methods). b Firing rate vs layer. For each neuron of each network, the firing rate was normalized by the average rate across the different groups. c Logistic map showing output rate (=firing rate in layer n + 1) versus input rate (=firing rate in layer n). The gray line has a slope of 1 and the red curve is a fit to the data. Dotted lines represent a sample trajectory starting at 40 Hz in layer 1 and ending at 13 Hz in layer 4. d Decoding accuracy quantified as the probability of correctly classifying firing rate in the first layer given the firing rate in the nth layer. Inset shows the decoding accuracy when 1, 2 or 3 neurons from layers 2–4 where used simultaneously to decode the firing rate in layer 1. Paired t-test statistic for the increase in decoding accuracy with the number of neurons: 1 vs 2: P = 1.1 × 10−5, 1 vs 3: P = 1.1 × 10−4, 2 vs 3: P = 0.012. e Same as in d for the channel capacity. Paired t-test statistic for the increase in channel capacity with the number of neurons: 1 vs 2: P = 0.002, 1 vs 3: P = 2.1 × 10−4, 2 vs 3: P = 0.004. Data are compiled from the same n = 16 dense networks of Fig. 2 and are presented as mean ± SEM. In d, e all data are shown as thin lines
