Kay, Kendrick
<p>Feb 13, 2010 10:02 AM
 
KENDRICK KAY ... his email is  knk@stanford.edu
 
I hear him lecture on Feb 12 at a FriSem  2010.  exclt.
 
worked with Jack Gallant  at Cal Berk.
 
has a 4 yr old dtr.
 
How is shape represented???   eg a bird??  ( I transcribed these notes in June... terrible transcript)
see his papers below
 
transforms from  retina to LGN to v1 v2 v4  IT
 
from  retinal detection to a percept ? in IT
 
receptor field models
 
Hubel Wiesel    v1  linear model of recp field.
 
how to do encoding and decoding of  images into percepts.
 
subjects in 4T fMRI   each voxel is  2mm 3
 
18 images;   do rigid body transform;
 
one photo every 4 seconds;   each flashes 3X
 
1750  images  each is difft;   120 images * 13 reps
 
get one beta per voxel.
 
v1  has 2000 voxels.
 
Gabor wavelet  do reconstruction from wavelets.
 
always fixate at center of image;
 
R = .3
 
does preserve retinotopy.
 
Gabor  accounts for  70% of the explainable variation at V1.
 
Haynes and Rees 2005  decoding;  Kamitane and Tony  2005
 
Can we build a decoder that predicts which image S is looking at?
 
Tom  Mitchelll 2008  celery  =>  food semantic model
 
Miyawaki  Neuron  2008  also did image reconstruc
 
Naselaris  Neuron 2009;
 
 </p>