CEBRA

CEBRA copied to clipboard

.. image:: https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1108e4f7-2b50-4b91-b558-e4c9ac908f53/cebra.png?format=2500w :width: 400

========

Welcome! This repo will host the official implementation for cebra, an algorithm to estimate C\ onsistent E\ m\ B\ eddings of high-dimensional R\ ecordings using A\ uxiliary variables. To receive updates on the code release, please :eyes: watch or :star: star this repository!

cebra is a self-supervised method for non-linear clustering that allows for label-informed time series analysis. It jointly uses behavioral and neural data in a hypothesis- or discovery-driven manner to produce consistent, high-performance latent spaces. The main 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.

• 🔗 Project Webpage: https://cebra.ai
• 📄 Preprint: Learnable latent embeddings for joint behavioral and neural analysis. Steffen Schneider*, Jin Hwa Lee* and Mackenzie Weygandt Mathis, arXiv:abs/2204.00673 <https://arxiv.org/abs/2204.00673>_