Next Level Processors

Neuromatch 2021 Deep Learning Project:

The Phenomenon: Stimuli (movie scenes, naturalistic, music etc) evoke attention and potentially similar brain activity across viewers.

The Data:

• Study Forrest Movie Transcript
• Study Forrest Eye Gaze and Neuroimaging
• Human Connectome Project Neuroimaging + Story Q/A

The Question: Can audience preference be linked to reliable neural representations of stimuli (scenes, naturalistic, music etc). Do more preferred/attention grabbing stimuli produce shared neural signals among audiences. In other words, can we predict sentiments (plus attention) based on the fMRI functional connectivity metric?

Approach & Goal To predict sentiments from functional connectivity in the brain. Use NLP derived sentiment metrics for audience preference!??! (Eye gaze/fixation/saccade for attention too!) And see if shared preferences evoke shared brain patterns.

Neuroscience: Can we use BOLD fMRI functional connectivity matrix to predict sentiments or stimuli features? Or the other way around?

In this repo:

• nl-procesors: code base for deep learning components
• data: datasets to play with... currently trying to reproduce existing work.