brainbodyfm-workshop.github.io/call-for-demos

This work was done with amazing collaborators:
Krystal Pan, @vinam.bsky.social, Ian Knight, Eva Dyer and @tyrellturing.bsky.social !

We find that the model’s latent representations carry meaningful information that reflect the anatomy and physiology of different regions and sub-types, even though it was never given any information about these distinctions!

Our results show benefits of scaling across multiple recordings and tasks. We also show that transfer to new datasets works really well, even when we're dealing with new tasks or new brain areas!

We train POYO+ on the @alleninstitute.bsky.social brain observatory dataset. That's 256 mice, 6 visual brain areas and 13 genetically defined cellular sub-types.

This is x10 more data than POYO-1.

POYO+ adds support for regression, classification, and segmentation tasks. It can be trained on multiple tasks at the same time!

We query POYO+ when decoding, meaning that it can be queried to decode any number of tasks, and these tasks can be different depending on the context.

POYO+ adds support for regular time series data through a value projection layer. We use it on calcium traces!

How is POYO+ different from POYO?
1. POYO+ is even more flexible, it supports more modalities and more tasks!
2. POYO+ is trained on 10x the data.
3. We provide analyses that reveal latent structure in the neural activity, modulated by brain areas, cell types, and tasks.

If I missed any relevant models, please let me know!