Cebra

Introducing CEBRA: Learnable Latent Embeddings for Joint Behavioural and Neural Analysis, an innovative machine-learning approach that bridges the gap between behavioral actions and neural activity—a cornerstone of neuroscience. As we capture ever-larger datasets of neural and behavioral data, CEBRA is designed to meet the growing demand for modeling neural dynamics during adaptive behaviors.

This groundbreaking tool offers a dual approach, allowing researchers to utilize both behavioral and neural data in a hypothesis-driven or a discovery-focused manner. CEBRA excels at generating powerful, consistent latent spaces that illuminate the connections between behavior and neural mechanisms. It can analyze both single and multi-session datasets for hypothesis testing or function label-free, providing unparalleled flexibility for researchers.

CEBRA is exceptionally versatile, adeptly handling calcium imaging and electrophysiological datasets across a range of tasks, including sensory and motor activities, whether they involve simple movements or complex behaviors, and across various species. One of its standout features is its ability to map spatial information, revealing intricate kinematic characteristics and enabling rapid, high-accuracy decoding of natural movies watched by subjects in the visual cortex.

Additionally, CEBRA’s ability to decode brain activity from the mouse visual cortex to reconstruct viewed videos underscores its potential for advancing our understanding of neural dynamics and behavior. Whether you're investigating sensory processing or exploring the nuances of motor control, CEBRA is poised to be a transformative tool in your research toolkit, propelling insights in neuroscience and behavioral studies to new heights.