dorsal/arxiv
View SchemaMaximally informative dimensions: Analyzing neural responses to natural signals
| Authors | Tatyana Sharpee, Nicole C. Rust, William Bialek |
|---|---|
| Categories | |
| ArXiv ID | physics/0208057 |
| URL | https://arxiv.org/abs/physics/0208057 |
Abstract
We propose a method that would allow for a rigorous statistical analysis of neural responses to natural stimuli, which are non-Gaussian and exhibit strong correlations. We have in mind a model in which neurons are selective for a small number of stimulus dimensions out of the high dimensional stimulus space, but within this subspace the responses can be arbitrarily nonlinear. Existing analysis methods are based on correlation functions between stimuli and responses, but these methods are guaranteed to work only in the case of Gaussian stimulus ensembles. As an alternative to correlation functions, we maximize the mutual information between the neural responses and projections of the stimulus onto low dimensional subspaces. The procedure can be done iteratively by increasing the dimensionality of this subspace. Those dimensions that allow the recovery of all of the information between spikes and the full unprojected stimuli describe the relevant subspace. If the dimensionality of the relevant subspace indeed is small, it becomes feasible to map out the neuron's input-output function even under fully natural stimulus conditions. These ideas are illustrated in simulations on model visual neurons responding to natural scenes.
{
"annotation_id": "ae38af00-d361-4d5a-ba16-16082ca3681f",
"date_created": "2026-03-02T18:00:39.175000Z",
"date_modified": "2026-03-02T18:00:39.175000Z",
"file_hash": "f1e1cc9d1bda9154134db3aa6e297d7d8aa1c86902869bd08b3d6de24253ddb9",
"private": false,
"record": {
"abstract": "We propose a method that would allow for a rigorous statistical analysis of\nneural responses to natural stimuli, which are non-Gaussian and exhibit strong\ncorrelations. We have in mind a model in which neurons are selective for a\nsmall number of stimulus dimensions out of the high dimensional stimulus space,\nbut within this subspace the responses can be arbitrarily nonlinear. Existing\nanalysis methods are based on correlation functions between stimuli and\nresponses, but these methods are guaranteed to work only in the case of\nGaussian stimulus ensembles. As an alternative to correlation functions, we\nmaximize the mutual information between the neural responses and projections of\nthe stimulus onto low dimensional subspaces. The procedure can be done\niteratively by increasing the dimensionality of this subspace. Those dimensions\nthat allow the recovery of all of the information between spikes and the full\nunprojected stimuli describe the relevant subspace. If the dimensionality of\nthe relevant subspace indeed is small, it becomes feasible to map out the\nneuron\u0027s input-output function even under fully natural stimulus conditions.\nThese ideas are illustrated in simulations on model visual neurons responding\nto natural scenes.",
"arxiv_id": "physics/0208057",
"authors": [
"Tatyana Sharpee",
"Nicole C. Rust",
"William Bialek"
],
"categories": [
"physics.bio-ph",
"physics.data-an",
"q-bio.NC"
],
"title": "Maximally informative dimensions: Analyzing neural responses to natural signals",
"url": "https://arxiv.org/abs/physics/0208057"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "6b3dfc66-f150-4d05-ba6d-7ce7de8dedad",
"id": "arXiv Dataset IDs",
"type": "Model",
"variant": "snapshot-2026-03-01",
"version": "0.1.0"
},
"user_id": 1000002
}