Fig 3. Excitatory and inhibitory population activity structure.

(A) Clustered and (B) non-clustered network shared dimensionality (top panels) and percent shared variance (bottom panels) with increasing neuron count with 10,000 trials. Dots (red: excitatory neurons, blue: inhibitory neurons) indicate the mean across five non-overlapping sets of sampled neurons and five non-overlapping sets of trials (25 sets total), and standard error bars are smaller than the dot size in all cases. To assess shared dimensionality and percent shared variance for a larger number of neurons, we grouped the five non-overlapping sets of 150 neurons from each neuron type population of each network into 750-neuron samplings. For 750-neuron samplings of the clustered network (10,000 trials), the excitatory population had a shared dimensionality of 23.95 ± 0.05 and percent shared variance of 91.43% ± 0.06%, while the inhibitory population had a shared dimensionality of 246.7 ± 2.5 and a percent shared variance of 79.56% ± 0.22%. For 750-neuron samplings of the non-clustered network (10,000 trials), the excitatory population had a shared dimensionality of 139.4 ± 3.8 and percent shared variance of 20.70% ± 0.43%, while the inhibitory population had a shared dimensionality of 347.2 ± 4.2 and a percent shared variance of 59.74% ± 0.47%. (C-D) same conventions as A and B, but for increasing trial count with 100 neurons.