[D] Question: Time-series decoding to non-temporal latent space?

Hello! I am a researcher in computational neuroscience, looking to apply some contemporary machine learning techniques to fMRI timeseries data. I have a collection of highly dimensional 4D fMRI timeseries data collected while subjects were observing naturalistic images from COCO at regular intervals. We currently have decoding models that take preprocessed "snapshots" of this timeseries data flattened into an activation pattern that is aggregated over the short period the image was being observed, and use some machine learning models to …

data decoding machine machine learning machinelearning machine learning techniques process series timeseries transformer