TRACE: Contrastive learning for multi-trial time-series data in neuroscience

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Schmors, Lisa, Gonschorek, Dominic, Böhm, Jan Niklas, Qiu, Yongrong, Zhou, Na, Kobak, Dmitry, Tolias, Andreas, Sinz, Fabian, Reimer, Jacob, Franke, Katrin, Damrich, Sebastian, Berens, Philipp
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912692423360512
author Schmors, Lisa
Gonschorek, Dominic
Böhm, Jan Niklas
Qiu, Yongrong
Zhou, Na
Kobak, Dmitry
Tolias, Andreas
Sinz, Fabian
Reimer, Jacob
Franke, Katrin
Damrich, Sebastian
Berens, Philipp
author_facet Schmors, Lisa
Gonschorek, Dominic
Böhm, Jan Niklas
Qiu, Yongrong
Zhou, Na
Kobak, Dmitry
Tolias, Andreas
Sinz, Fabian
Reimer, Jacob
Franke, Katrin
Damrich, Sebastian
Berens, Philipp
contents Modern neural recording techniques such as two-photon imaging or Neuropixel probes allow to acquire vast time-series datasets with responses of hundreds or thousands of neurons. Contrastive learning is a powerful self-supervised framework for learning representations of complex datasets. Existing applications for neural time series rely on generic data augmentations and do not exploit the multi-trial data structure inherent in many neural datasets. Here we present TRACE, a new contrastive learning framework that averages across different subsets of trials to generate positive pairs. TRACE allows to directly learn a two-dimensional embedding, combining ideas from contrastive learning and neighbor embeddings. We show that TRACE outperforms other methods, resolving fine response differences in simulated data. Further, using in vivo recordings, we show that the representations learned by TRACE capture both biologically relevant continuous variation, cell-type-related cluster structure, and can assist data quality control.
format Preprint
id arxiv_https___arxiv_org_abs_2506_04906
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TRACE: Contrastive learning for multi-trial time-series data in neuroscience
Schmors, Lisa
Gonschorek, Dominic
Böhm, Jan Niklas
Qiu, Yongrong
Zhou, Na
Kobak, Dmitry
Tolias, Andreas
Sinz, Fabian
Reimer, Jacob
Franke, Katrin
Damrich, Sebastian
Berens, Philipp
Neurons and Cognition
Modern neural recording techniques such as two-photon imaging or Neuropixel probes allow to acquire vast time-series datasets with responses of hundreds or thousands of neurons. Contrastive learning is a powerful self-supervised framework for learning representations of complex datasets. Existing applications for neural time series rely on generic data augmentations and do not exploit the multi-trial data structure inherent in many neural datasets. Here we present TRACE, a new contrastive learning framework that averages across different subsets of trials to generate positive pairs. TRACE allows to directly learn a two-dimensional embedding, combining ideas from contrastive learning and neighbor embeddings. We show that TRACE outperforms other methods, resolving fine response differences in simulated data. Further, using in vivo recordings, we show that the representations learned by TRACE capture both biologically relevant continuous variation, cell-type-related cluster structure, and can assist data quality control.
title TRACE: Contrastive learning for multi-trial time-series data in neuroscience
topic Neurons and Cognition
url https://arxiv.org/abs/2506.04906