Figure 5
Estimating the null and potent space of neural communication. (a) The Frobenius norm of the null and potent space of connections W0 intersected at a location where approximately half of all modes are null (N = 100). (b) The ratio of null and potent neural modes (Eq. (3)) provided an estimation of the proportion of null modes in the model, but also generated unwanted nonlinearity. (c) Correcting nonlinearity by adding a term to the ratio (Eq. (4)) led to an overshoot in prediction. (d) The overshoot in prediction was corrected by scaling the estimation with a factor inversely proportional to the square root of N (Eq. (5)). (e) Root mean square difference (RMSD) between actual and estimated null ratios across different durations of simulation. Vertical dashed line: one second of activity.
