Efficient Identification of Assembly Neurons within Massively Parallel Spike Trains

Figure 2

Analysis results for datasets representing the settings Indep, SIP, MIP, and mSIP using 𝑡^CPC₁ (a) and 𝑡^CSF₁ (b). All sets are composed of 𝑁=100 neurons. In Indep, neurons 1–10 (𝑀=10) have a higher firing rate (𝜆=50 Hz) than the rest of the neurons (𝜆=20 Hz). In SIP, MIP and mSIP the first 𝑀=10 neurons are involved in synchronous events, the rest are independent. All neurons have the same firing rate, 𝜆=20 Hz. The coincidence firing rate of the assemblies is 𝜆_𝑐=5 Hz. The synchronous events in MIP are generated with a copy probability of 𝜖=0.8. In each panel, box plots show the distribution (wide box: 5% to 95%, narrow box: 1% to 99%, whiskers: minimum to maximum) of the shuffling results (𝑠=10⁵ runs) for each neuron 𝑖 (ids ordered along the 𝑥-axis). Beyond the extent of the whiskers, results are significant on a 1/𝑠 level. The test statistic values 𝑡^CPC₁ and 𝑡^CSF₁ obtained on the actual data are shown as bow-ties.

(a)

(b)