Inhibitory neural networks are found to encode high volumes of information through delayed inhibition. We show that inhibition delay increases storage capacity through a Stirling transform of the minimum capacity which stabilizes locally coherent oscillations. We obtain both the exact and asymptotic formulas for the total number of dynamic attractors. Our results predict a (ln2)-N-fold increase in capacity for an N-neuron network and demonstrate high-density associative memories which host a maximum number of oscillations in analog neural devices.
ASJC Scopus subject areas
- Statistical and Nonlinear Physics
- Statistics and Probability
- Condensed Matter Physics