CEBRA is a machine learning tool that uses non-linear techniques to create consistent and high-performance latent spaces from joint behavioural and neural data recorded simultaneously.
This tool allows for the mapping of behavioural actions to neural activity for a better understanding of neural dynamics during adaptive behaviours and can reveal underlying correlates of behaviour.
CEBRA creates neural latent embeddings that can be used for both hypothesis testing and discovery-driven analysis. The tool has been validated for its accuracy and efficacy on calcium and electrophysiology datasets, as well as across sensory and motor tasks and in simple or complex behaviours across species.
It can be used with single or multi-session datasets and can be used label-free. CEBRA can map and uncover complex kinematic features, produce consistent latent spaces across 2-photon and Neuropixels data, and provide rapid high-accuracy decoding of natural movies from visual cortex.
The tool’s code is available on GitHub and the pre-print is available on arxiv.org. CEBRA is a useful tool for neuroscientists who want to analyze and decode behavioural and neural data to reveal underlying neural representations.