Preintegration lateral inhibition enhances unsupervised learning
Neural Computation
Selectively grouping neurons in recurrent networks of lateral inhibition
Neural Computation
Neural Computation
Selectivity and Stability via Dendritic Nonlinearity
Neural Computation
Hi-index | 0.01 |
Recent experiments revealed that there exist electrophysiologically, anatomically, and functionally distinct two classes of GABAergic interneurons in the cerebral cortex, fast spiking (FS) cells, and non-FS cells. We propose a network model of cortical local circuits including dendritic inhibition, which is the anatomical hallmark of the non-FS cells. While conventional lateral inhibition models always converge to winner-take-all states if the self-excitation is strong, our model does so only for appropriate inputs, but otherwise converges to another states, in which all the neurons have little activities, even if the self-excitation is strong enough to keep the winner's activity after the extinction of the inputs.