CEBRA

CEBRA is a machine-learning method for compressing time series data to reveal hidden structures, particularly in simultaneously recorded behavioral and neural data. It uses a self-supervised learning algorithm to obtain interpretable, consistent embeddings of high-dimensional recordings using auxiliary variables.

Key Features

• Jointly uses behavioural and neural data in a supervised (hypothesis-driven) or self-supervised (discovery-driven) manner
• Produces consistent and high-performance latent spaces
• Can be used for decoding neural activity to reconstruct viewed videos, decode trajectories, and decode position during navigation
• Works with calcium and electrophysiology datasets across sensory and motor tasks
• Supports single and multi-session datasets for hypothesis testing or label-free use

Use Cases

• Decoding activity from the visual cortex of the mouse brain to reconstruct a viewed video
• Decoding trajectories from the sensorimotor cortex of primates
• Decoding position during navigation (e.g., rat hippocampus data)
• Mapping of space and uncovering complex kinematic features
• Producing consistent latent spaces across 2-photon and Neuropixels data

Who It's For

Neuroscience researchers and computational scientists working with neural and behavioral data.