<html>
Monday March 1, 2004<br>
Noon E25-401<br><br>
<font face="Times New Roman TUR, Times"><b>Gabriel Kreiman<br>
</b>Reading out information from the macaque inferior temporal
cortex<br>
Poggio Lab<br><br>
Primates have an extraordinary capacity to visually recognize objects.
This ability depends on the function of the inferior temporal (IT) cortex
in the macaque brain. Neurons in inferior temporal cortex are selective
for complex stimuli including faces and objects. How well can we decode
the activity in IT to be able to interpret which visual stimuli were
presented to the monkey? In collaboration with Chou Hung in Jim DiCarlo?s
lab, we are recording neuronal activity while a monkey is presented with
a set of 78 different visual stimuli. We are focusing on three different
sources of neuronal activity multi-unit activity (MUA) from single
electrodes, single-unit activity (SUA) and local field potentials (LFP).
The neuronal activity is fed to a classifier (Fisher linear discriminant
of Support Vector Machine) to learn the mapping from neuronal responses
to visual stimuli. We will describe the algorithms used to predict the
stimuli based on the neuronal activity and their dependency on different
possible inputs to the classifiers. We will report how the performance of
the classifier depends on the type of input (MUA, LFP), on the time
interval, on the timing of spikes and on the number of recording sites
(assuming independence). This work is a collaboration between Alec
Shkolnik, Chou Hung, Gabriel Kreiman, Jim DiCarlo and Tommy
Poggio.<br><br>
</font>Pizza will be served!<br><br>
All faculty, postdocs and students are invited to attend.<br><br>
Interested in giving a Brain Lunch talk? Dates still available for
this semester. Send me an email - ASAP<br><br>
For questions about Brain Lunch, contact Monica Linden ( mlinden@mit.edu
).<br><br>
See you on Monday,<br>
Monica Linden<br>
Brain Lunch Coordinator </html>