Cebra is an advanced machine learning tool that utilizes non-linear techniques to generate reliable and high-performing latent spaces from joint behavioral and neural data collected simultaneously.

Main Features:

• Neural Latent Embeddings: Utilize for hypothesis testing and discovery-oriented analysis.
• Validated Accuracy: Effectiveness has been demonstrated on various datasets related to calcium and electrophysiology, sensory and motor tasks, as well as simple or complex behaviors across different species.
• Multi-session and Label-free: Compatible with single or multi-session datasets and does not require labels.
• High-accuracy Decoding: Enables quick decoding of natural movies derived from the visual cortex.
• Code Availability: Access the tool's source code on GitHub and read the pre-print version on arxiv.org.

Possible Applications:

• Analyze and decode behavioral and neural data to uncover underlying neural representations.

• Map and reveal intricate kinematic features in neuroscience research.

• Generate consistent latent spaces across different data types and experiments.

Cebra is an invaluable resource for neuroscientists aiming to analyze and decode behavioral and neural data, thereby enhancing their understanding of the neural representations implicated in adaptive behaviors.