William Ngiam | 严祥全
Lecturer at the University of Adelaide | Perception, Attention, Learning and Memory Lab (https://palm-lab.github.io) | Open Practices Editor at Attention, Perception, & Psychophysics | Science reform through ReproducibiliTea | http://williamngiam.github.io
- A fascinating discussion; highly recommend for #workingmemory peeps. I'm more cog than I am neuro, so I don't know if I got it all, but good to reflect on "readout". My take is that a lot that is discarded in our models – such as preparatory dynamics (in PFC? hippocampal?) before encoding into WM.
-
View full threadHow do you get MLR2.0 to resolve/convolve the various layers? I'm also trying to *keep in mind* how we models use argmax on SDT (TCC/interference model) or correlation with preferred signal (neural resource). Seems like a lot if happening to build the representation (and then read out).
- (I wish I could teleport to VSS to have this conversation over a meal, but I will settle for a virtual chat if you've got the time! Or I can cause a ruckus during WMS).
