Efficient Identification of Assembly Neurons within Massively Parallel Spike Trains

Figure 3

False positive and false negative rate as a function of assembly size 𝑀. Each panel shows the FN rate as bar plots (positive 𝑦-axis) and FP rate (negative 𝑦-axis, and inset) for each type of dataset (a): SIP, 𝑁=100; (b) MIP, 𝑁=100, 𝜖=0.8; (c) SIP, 𝑁=1000; (d) MIP, 𝑁=1000, 𝜖=0.8. Each panel contains a sets of grouped bars (colors indicate the results for the different test statistics 𝑡^CPC₁, 𝑡^CPC₃, 𝑡^CPC₁, 𝑡^CSF₃; see legend) for each 𝑀. Each bar represents the mean of 10 realizations, and the error bar indicates one standard deviation. All datasets have stationary firing rates (all neurons 𝜆=20 Hz), a stationary coincidence rate of 𝜆_𝑐=1 Hz, and a duration of 𝑘=10000 time steps of ℎ=1 milliseconds. Significance level is 1%. For generation of the surrogates (5000 repetitions), spikes are shuffled with uniform distribution in time.

(a)

(b)

(c)

(d)