Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience

Emily L. Mackevicius, Andrew H. Bahle, Alex H. Williams, Shijie Gu, Natalia I. Denisenko, Mark S Goldman, Michale S. Fee

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Identifying low-dimensional features that describe large-scale neural recordings is a major challenge in neuroscience. Repeated temporal patterns (sequences) are thought to be a salient feature of neural dynamics, but are not succinctly captured by traditional dimensionality reduction techniques. Here, we describe a software toolbox-called seqNMF-with new methods for extracting informative, non-redundant, sequences from high-dimensional neural data, testing the significance of these extracted patterns, and assessing the prevalence of sequential structure in data. We test these methods on simulated data under multiple noise conditions, and on several real neural and behavioral datas. In hippocampal data, seqNMF identifies neural sequences that match those calculated manually by reference to behavioral events. In songbird data, seqNMF discovers neural sequences in untutored birds that lack stereotyped songs. Thus, by identifying temporal structure directly from neural data, seqNMF enables dissection of complex neural circuits without relying on temporal references from stimuli or behavioral outputs.

Original languageEnglish (US)
JournaleLife
Volume8
DOIs
StatePublished - Feb 5 2019

Fingerprint

Dissection
Birds
Neurosciences
Networks (circuits)
Testing
Songbirds
Music
Noise
Software
Datasets

Keywords

  • matrix factorization
  • neuroscience
  • rat
  • sequence
  • unsupervised
  • Zebra finch

ASJC Scopus subject areas

  • Neuroscience(all)
  • Biochemistry, Genetics and Molecular Biology(all)
  • Immunology and Microbiology(all)

Cite this

Mackevicius, E. L., Bahle, A. H., Williams, A. H., Gu, S., Denisenko, N. I., Goldman, M. S., & Fee, M. S. (2019). Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience. eLife, 8. https://doi.org/10.7554/eLife.38471

Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience. / Mackevicius, Emily L.; Bahle, Andrew H.; Williams, Alex H.; Gu, Shijie; Denisenko, Natalia I.; Goldman, Mark S; Fee, Michale S.

In: eLife, Vol. 8, 05.02.2019.

Research output: Contribution to journalArticle

Mackevicius, EL, Bahle, AH, Williams, AH, Gu, S, Denisenko, NI, Goldman, MS & Fee, MS 2019, 'Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience', eLife, vol. 8. https://doi.org/10.7554/eLife.38471
Mackevicius, Emily L. ; Bahle, Andrew H. ; Williams, Alex H. ; Gu, Shijie ; Denisenko, Natalia I. ; Goldman, Mark S ; Fee, Michale S. / Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience. In: eLife. 2019 ; Vol. 8.
@article{c068296c4ec94773af862ab281c8c71c,
title = "Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience",
abstract = "Identifying low-dimensional features that describe large-scale neural recordings is a major challenge in neuroscience. Repeated temporal patterns (sequences) are thought to be a salient feature of neural dynamics, but are not succinctly captured by traditional dimensionality reduction techniques. Here, we describe a software toolbox-called seqNMF-with new methods for extracting informative, non-redundant, sequences from high-dimensional neural data, testing the significance of these extracted patterns, and assessing the prevalence of sequential structure in data. We test these methods on simulated data under multiple noise conditions, and on several real neural and behavioral datas. In hippocampal data, seqNMF identifies neural sequences that match those calculated manually by reference to behavioral events. In songbird data, seqNMF discovers neural sequences in untutored birds that lack stereotyped songs. Thus, by identifying temporal structure directly from neural data, seqNMF enables dissection of complex neural circuits without relying on temporal references from stimuli or behavioral outputs.",
keywords = "matrix factorization, neuroscience, rat, sequence, unsupervised, Zebra finch",
author = "Mackevicius, {Emily L.} and Bahle, {Andrew H.} and Williams, {Alex H.} and Shijie Gu and Denisenko, {Natalia I.} and Goldman, {Mark S} and Fee, {Michale S.}",
year = "2019",
month = "2",
day = "5",
doi = "10.7554/eLife.38471",
language = "English (US)",
volume = "8",
journal = "eLife",
issn = "2050-084X",
publisher = "eLife Sciences Publications",

}

TY - JOUR

T1 - Unsupervised discovery of temporal sequences in high-dimensional datasets, with applications to neuroscience

AU - Mackevicius, Emily L.

AU - Bahle, Andrew H.

AU - Williams, Alex H.

AU - Gu, Shijie

AU - Denisenko, Natalia I.

AU - Goldman, Mark S

AU - Fee, Michale S.

PY - 2019/2/5

Y1 - 2019/2/5

N2 - Identifying low-dimensional features that describe large-scale neural recordings is a major challenge in neuroscience. Repeated temporal patterns (sequences) are thought to be a salient feature of neural dynamics, but are not succinctly captured by traditional dimensionality reduction techniques. Here, we describe a software toolbox-called seqNMF-with new methods for extracting informative, non-redundant, sequences from high-dimensional neural data, testing the significance of these extracted patterns, and assessing the prevalence of sequential structure in data. We test these methods on simulated data under multiple noise conditions, and on several real neural and behavioral datas. In hippocampal data, seqNMF identifies neural sequences that match those calculated manually by reference to behavioral events. In songbird data, seqNMF discovers neural sequences in untutored birds that lack stereotyped songs. Thus, by identifying temporal structure directly from neural data, seqNMF enables dissection of complex neural circuits without relying on temporal references from stimuli or behavioral outputs.

AB - Identifying low-dimensional features that describe large-scale neural recordings is a major challenge in neuroscience. Repeated temporal patterns (sequences) are thought to be a salient feature of neural dynamics, but are not succinctly captured by traditional dimensionality reduction techniques. Here, we describe a software toolbox-called seqNMF-with new methods for extracting informative, non-redundant, sequences from high-dimensional neural data, testing the significance of these extracted patterns, and assessing the prevalence of sequential structure in data. We test these methods on simulated data under multiple noise conditions, and on several real neural and behavioral datas. In hippocampal data, seqNMF identifies neural sequences that match those calculated manually by reference to behavioral events. In songbird data, seqNMF discovers neural sequences in untutored birds that lack stereotyped songs. Thus, by identifying temporal structure directly from neural data, seqNMF enables dissection of complex neural circuits without relying on temporal references from stimuli or behavioral outputs.

KW - matrix factorization

KW - neuroscience

KW - rat

KW - sequence

KW - unsupervised

KW - Zebra finch

UR - http://www.scopus.com/inward/record.url?scp=85061054035&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061054035&partnerID=8YFLogxK

U2 - 10.7554/eLife.38471

DO - 10.7554/eLife.38471

M3 - Article

C2 - 30719973

AN - SCOPUS:85061054035

VL - 8

JO - eLife

JF - eLife

SN - 2050-084X

ER -