HIDA Lecture: Large-Scale Brain Decoding

Abstract:

Decoding information from neural activity can be important for many neuroscience applications, including basic research into sensorimotor processing, clinical diagnosis, and brain-computer interfaces. Brain data is too complex for traditional statistical techniques to provide much traction in decoding, and there is general agreement that deep learning approaches are necessary. However, deep learning requires large-scale datasets, which presents a challenge for neural activity data, because neural data is highly heterogeneous, with each recording involving different subjects, modalities, species, brain regions, stimuli, behaviour, and health states. How can we train a single model across such diverse data?

The lecture introduces new techniques developed for training neural decoding models across diverse datasets. Using novel tokenization approaches, Blake Richards demonstrates the feasibility of building large-scale, transformer-based neural decoding models that effectively integrate data from different subjects, stimuli or behaviors, brain regions, cell types, and even species. Their findings reveal that model performance improves as data diversity increases, suggesting that scaling neural decoding models is achievable when the system is specifically designed to leverage the richness of neural datasets.

Register now!