Learning low-dimensional generalizable naturalfeatures from retina using a U-net
Siwei Wang, Benjamin D. Hoshal, Elizabeth A. de Laittre, Olivier Marre, Michael Berry, Stephanie Palmer
The paper uses a U-net architecture to learn a low-dimensional, generalizable representation of time in natural scenes from retinal ganglion cell activity, achieving 17 ms resolution across different movies.
neural information processing systems · Jan 1, 2022Peer-reviewedNeuroscience