Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires

Increases in the scale and complexity of behavioral data pose an increasing challenge for data analysis. A common strategy involves replacing entire behaviors with small numbers of handpicked, domain-specific features, but this approach suffers from several crucial limitations. For example, handpick...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Jack Goffinet, Samuel Brudner, Richard Mooney, John Pearson
Formato:	article
Lenguaje:	EN
Publicado:	eLife Sciences Publications Ltd 2021
Materias:	zebra finch autoencoder statistics Medicine R Science Q Biology (General) QH301-705.5
Acceso en línea:	https://doaj.org/article/55763c7b25704292b75e302878ae32a4
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:55763c7b25704292b75e302878ae32a4
record_format	dspace
spelling	oai:doaj.org-article:55763c7b25704292b75e302878ae32a42021-11-16T14:23:35ZLow-dimensional learned feature spaces quantify individual and group differences in vocal repertoires10.7554/eLife.678552050-084Xe67855https://doaj.org/article/55763c7b25704292b75e302878ae32a42021-05-01T00:00:00Zhttps://elifesciences.org/articles/67855https://doaj.org/toc/2050-084XIncreases in the scale and complexity of behavioral data pose an increasing challenge for data analysis. A common strategy involves replacing entire behaviors with small numbers of handpicked, domain-specific features, but this approach suffers from several crucial limitations. For example, handpicked features may miss important dimensions of variability, and correlations among them complicate statistical testing. Here, by contrast, we apply the variational autoencoder (VAE), an unsupervised learning method, to learn features directly from data and quantify the vocal behavior of two model species: the laboratory mouse and the zebra finch. The VAE converges on a parsimonious representation that outperforms handpicked features on a variety of common analysis tasks, enables the measurement of moment-by-moment vocal variability on the timescale of tens of milliseconds in the zebra finch, provides strong evidence that mouse ultrasonic vocalizations do not cluster as is commonly believed, and captures the similarity of tutor and pupil birdsong with qualitatively higher fidelity than previous approaches. In all, we demonstrate the utility of modern unsupervised learning approaches to the quantification of complex and high-dimensional vocal behavior.Jack GoffinetSamuel BrudnerRichard MooneyJohn PearsoneLife Sciences Publications Ltdarticlezebra finchautoencoderstatisticsMedicineRScienceQBiology (General)QH301-705.5ENeLife, Vol 10 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	zebra finch autoencoder statistics Medicine R Science Q Biology (General) QH301-705.5
spellingShingle	zebra finch autoencoder statistics Medicine R Science Q Biology (General) QH301-705.5 Jack Goffinet Samuel Brudner Richard Mooney John Pearson Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
description	Increases in the scale and complexity of behavioral data pose an increasing challenge for data analysis. A common strategy involves replacing entire behaviors with small numbers of handpicked, domain-specific features, but this approach suffers from several crucial limitations. For example, handpicked features may miss important dimensions of variability, and correlations among them complicate statistical testing. Here, by contrast, we apply the variational autoencoder (VAE), an unsupervised learning method, to learn features directly from data and quantify the vocal behavior of two model species: the laboratory mouse and the zebra finch. The VAE converges on a parsimonious representation that outperforms handpicked features on a variety of common analysis tasks, enables the measurement of moment-by-moment vocal variability on the timescale of tens of milliseconds in the zebra finch, provides strong evidence that mouse ultrasonic vocalizations do not cluster as is commonly believed, and captures the similarity of tutor and pupil birdsong with qualitatively higher fidelity than previous approaches. In all, we demonstrate the utility of modern unsupervised learning approaches to the quantification of complex and high-dimensional vocal behavior.
format	article
author	Jack Goffinet Samuel Brudner Richard Mooney John Pearson
author_facet	Jack Goffinet Samuel Brudner Richard Mooney John Pearson
author_sort	Jack Goffinet
title	Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
title_short	Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
title_full	Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
title_fullStr	Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
title_full_unstemmed	Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
title_sort	low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires
publisher	eLife Sciences Publications Ltd
publishDate	2021
url	https://doaj.org/article/55763c7b25704292b75e302878ae32a4
work_keys_str_mv	AT jackgoffinet lowdimensionallearnedfeaturespacesquantifyindividualandgroupdifferencesinvocalrepertoires AT samuelbrudner lowdimensionallearnedfeaturespacesquantifyindividualandgroupdifferencesinvocalrepertoires AT richardmooney lowdimensionallearnedfeaturespacesquantifyindividualandgroupdifferencesinvocalrepertoires AT johnpearson lowdimensionallearnedfeaturespacesquantifyindividualandgroupdifferencesinvocalrepertoires
_version_	1718426358497935360

Low-dimensional learned feature spaces quantify individual and group differences in vocal repertoires

Ejemplares similares