Figure 8. Storage capacity.

(a) Storage capacity of a diluted network: dependence on γ and $1/L$ (represented as $lo{g}_{10}(1/L)$). (b) Storage capacity of a fully connected network: non monotonic dependence of the capacity on γ. Retrieval / no retrieval phase transition for different values of γ, obtained from simulations with $N=1000$, $N_S=10$ and $L=10$, for 1D manifolds. Error bars show the standard error of the observed proportion of successful retrievals. The non-monotonic dependence of the capacity from γ can be appreciated here: the transition point moves toward the right with increasing γ up to $\gamma \sim 1$, then back to the left. (c) Storage capacity of a fully connected network as a function of map sparsity $1/L$ and asymmetry strength γ, for a one-dimensional and a two-dimensional dynamic continuous attractor.