Skip to main content
. 2011 Jun 15;6(6):e20996. doi: 10.1371/journal.pone.0020996

Figure 2. PCs do not always isolate the activity of different cell assemblies.

Figure 2

(A) Top panel shows a binned spike activity matrix with 20 neurons (modeled as Poissonian processes) and 8000 time bins. Two cell assemblies were simulated in the network, each having four neurons (Assembly 1 neurons: #6, #7, #8, #9; Assembly 2 neurons: #12, #13, #14, #15). Neurons in the same assembly were set to fire together six times above their mean firing rate at 0.5% of the bins. Bottom panels show the estimated time course of ensemble activity obtained by the projection of the binned spike activity using the projector operator defined as the outer product of the PCs (see Methods). Note that PC1 marks the activations of Assembly 2, and PC2 marks the activations of Assembly 1. (B) Same as A, but with assemblies sharing neurons (Assembly 1 neurons: #5, #6, #7, #8, #9; Assembly 2 neurons: #8, #9, #10, #11, #12). Note that for this example this framework fails to isolate the activity of individual assemblies.