dorsal/arxiv
View SchemaFine Discrimination of Analog Patterns by Nonlinear Dendritic Inhibition
| Authors | Kenji Morita, Kazuyuki Aihara |
|---|---|
| Categories | |
| ArXiv ID | q-bio/0402035 |
| URL | https://arxiv.org/abs/q-bio/0402035 |
Abstract
Recent experiments revealed that a certain class of inhibitory neurons in the cerebral cortex make synapses not onto cell bodies but at distal parts of dendrites of the target neurons, mediating highly nonlinear dendritic inhibition. We propose a novel form of competitive neural network model that realizes such dendritic inhibition. Contrary to the conventional lateral inhibition in neural networks, our dendritic inhibition models don't always show winner-take-all behaviors; instead, they converge to "I don't know" states when unknown input patterns are presented. We derive reduced two-dimensional dynamics for the network, showing that a drastic shift of the fixed point from a winner-take-all state to an "I don't know" state occurs in accordance with the increase in noise added to the stored patterns. By preventing misrecognition in such a way, dendritic inhibition networks achieve fine pattern discrimination, which could be one of the basic computations by inhibitory connected recurrent neural networks in the brain.
{
"annotation_id": "87122c81-e501-4137-ba4e-5e2caa37b0a4",
"date_created": "2026-03-02T18:01:32.179000Z",
"date_modified": "2026-03-02T18:01:32.179000Z",
"file_hash": "50b01885ecefa0460e1b2492a5db2d8264670a0570356dfc438ac01b6559233d",
"private": false,
"record": {
"abstract": "Recent experiments revealed that a certain class of inhibitory neurons in the\ncerebral cortex make synapses not onto cell bodies but at distal parts of\ndendrites of the target neurons, mediating highly nonlinear dendritic\ninhibition. We propose a novel form of competitive neural network model that\nrealizes such dendritic inhibition. Contrary to the conventional lateral\ninhibition in neural networks, our dendritic inhibition models don\u0027t always\nshow winner-take-all behaviors; instead, they converge to \"I don\u0027t know\" states\nwhen unknown input patterns are presented. We derive reduced two-dimensional\ndynamics for the network, showing that a drastic shift of the fixed point from\na winner-take-all state to an \"I don\u0027t know\" state occurs in accordance with\nthe increase in noise added to the stored patterns. By preventing\nmisrecognition in such a way, dendritic inhibition networks achieve fine\npattern discrimination, which could be one of the basic computations by\ninhibitory connected recurrent neural networks in the brain.",
"arxiv_id": "q-bio/0402035",
"authors": [
"Kenji Morita",
"Kazuyuki Aihara"
],
"categories": [
"q-bio.NC"
],
"title": "Fine Discrimination of Analog Patterns by Nonlinear Dendritic Inhibition",
"url": "https://arxiv.org/abs/q-bio/0402035"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "4d9bb897-0d75-4a8b-9c72-721cd8b834ce",
"id": "arXiv Dataset IDs",
"type": "Model",
"variant": "snapshot-2026-03-01",
"version": "0.1.0"
},
"user_id": 1000002
}