Cebra

Cebra
Pricing: Free
Type: Research
Starts: $0/m

CEBRA is a machine-learning method that compresses time series data, revealing hidden structures in the variability. It is effective for analyzing behavioral and neural data, and can even reconstruct viewed videos from activity in the visual cortex of the mouse brain.

­čžá Use cases

  • Behavioral and Neural Analysis:┬áCEBRA can be used to map behavioral actions to neural activity, uncovering neural dynamics during adaptive behaviors.
  • Neural Latent Embeddings:┬áCEBRA generates consistent and high-performance latent spaces that reveal underlying correlates of behavior, allowing for decoding and hypothesis testing.
  • Calcium and Electrophysiology Datasets:┬áCEBRA is accurate and useful for analyzing datasets from calcium imaging and electrophysiology across sensory and motor tasks.
  • Species and Task Flexibility:┬áCEBRA can be applied to both simple and complex behaviors across different species, making it versatile for studying various experimental paradigms.
  • Space Mapping:┬áCEBRA can be used to map space and uncover complex kinematic features.
  • 2-photon and Neuropixels Data:┬áCEBRA produces consistent latent spaces across different types of neural recordings, such as 2-photon imaging and Neuropixels.
  • Decoding Natural Movies:┬áCEBRA enables rapid and high-accuracy decoding of natural movies from the visual cortex.

­čžá Features

  • Consistency Metric:┬áConsistency serves as a metric for uncovering meaningful differences in the latent spaces generated by CEBRA.
  • Single and Multi-Session Analysis:┬áCEBRA can leverage single and multi-session datasets for hypothesis testing or can be used in a label-free manner.
  • Open-Source Implementation:┬áThe official implementation of the CEBRA algorithm is available on GitHub, where updates and releases can be found. The repository can be watched and starred for notifications.
  • Collaboration Opportunities:┬áFor collaborations, the CEBRA team can be contacted via email.
  • Citation:┬áThe paper should be cited as:
    • Schneider, S., Lee, J. H., & Mathis, M. W. (2023). “Learnable latent embeddings for joint behavioral and neural analysis.” Nature. May 3

Sign In

Register

Reset Password

Please enter your username or email address, you will receive a link to create a new password via email.