Welcome! 👋

CEBRA is a library for estimating Consistent EmBeddings of high-dimensional Recordings using Auxiliary variables. It contains self-supervised learning algorithms implemented in PyTorch, and has support for a variety of different datasets common in biology and neuroscience.

To receive updates on code releases, please 👀 watch or ⭐️ star this repository!

cebra is a patented self-supervised method for non-linear clustering that allows for label-informed time series analysis. It can jointly use behavioral and neural data in a hypothesis- or discovery-driven manner to produce consistent, high-performance latent spaces. While it is not specific to neural and behavioral data, this is the first domain we used the tool in. This application case is to obtain a consistent representation of latent variables driving activity and behavior, improving decoding accuracy of behavioral variables over standard supervised learning, and obtaining embeddings which are robust to domain shifts.

References

• 📄 Publication April 2025: Time-series attribution maps with regularized contrastive learning. Steffen Schneider, Rodrigo González Laiz, Anastasiia Filipova, Markus Frey, Mackenzie Weygandt Mathis. AISTATS 2025.

• 📄 Publication May 2023: Learnable latent embeddings for joint behavioural and neural analysis. Steffen Schneider*, Jin Hwa Lee* and Mackenzie Weygandt Mathis. Nature 2023.

• 📄 Preprint April 2022: Learnable latent embeddings for joint behavioral and neural analysis. Steffen Schneider*, Jin Hwa Lee* and Mackenzie Weygandt Mathis

Patent Information

• Dimensionality reduction of time-series data, and systems and devices that use the resultant embeddings. Steffen Schneider* & Mackenzie Weygandt Mathis*. Awarded Dec 2025. Please contact the TTO office at EPFL for licensing.

License

• Since version 0.4.0, CEBRA is open source software under an Apache 2.0 license.
• Prior versions 0.1.0 to 0.3.1 were released for academic use only (please read the license file).