Computing and stability in cortical networks
Neural Computation
2008 Special Issue: The state of MIIND
Neural Networks
| Hi-index | 0.00 |
Delay activity (DA) is the increased firing rate of a cortical population, which persists when the stimulus that induced it is removed. It is believed to be the neural substrate for working memory, and as such highly relevant for theories of cognition. The cortex is highly recurrent, mainly excitatory, and finding stable attractors for DA at low firing rates for realistic neuronal parameters has proven to be hard. Most models for DA use recurrent excitation. Here a model with recurrent disinhibition is presented, which is manifestly stable. This model requires a cortical circuit that is slightly more complex than circuits in models using recurrent excitation, but circuits of comparable complexity have been found in cortex. Since delay attractors can not be observed directly, it is important to consider all theoretical possibilities.