Twitter | Pretraživanje | |
Il Memming Park
Fresh on . Latent trajectory analysis of MT ensemble spike trains reveals mis-aligned subspaces of stimulus and decision codes. Collaboration with Huk lab.
Reply Retweet Označi sa "sviđa mi se" More
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
Trial-to-trial variability of single unit spike count in monkey brain area MT is correlated with the animal’s perceptual report in visual motion discrimination tasks [Britten et al. 1996]. Where does this choice-correlation originate from?
Reply Retweet Označi sa "sviđa mi se"
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
Our ensemble analysis revealed that 1-D visual motion is encoded as 1-D subspace at the population level (below). Does the choice-correlation also live solely in this subspace? Maybe it should be, since MT is a “sensory cortex”, and the monkey is nearly “optimal”.
Reply Retweet Označi sa "sviđa mi se"
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
But it could also be due to feedback, i.e., noisy temporal integration of sensory evidence can be contributing to the apparent variability in MT. Depending on the alignment between the sensory-coding subspace and the choice-correlation, there are several possibilities:
Reply Retweet Označi sa "sviđa mi se"
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
Our analysis suggests that a large portion of choice-correlated variability in MT is in a subspace NOT aligned with the sensory encoding and lives in the *null-space*!
Reply Retweet Označi sa "sviđa mi se"
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
Furthermore, the choice-correlated variability in the null-space appears much later during the trial, suggesting that this is feedback information from the decision-making process! This information could be useful for learning and adaptation.
Reply Retweet Označi sa "sviđa mi se"
Il Memming Park 10. sij
Odgovor korisniku/ci @biorxiv @yuanz271 @jcbyts
Methods: all analysis of latent space was done using vLGP . MT ensemble was well described by 4-D latent space. (also 's brilliant work)
Reply Retweet Označi sa "sviđa mi se"