Matching categorical object representation in the inferior temporal cortex of man and monkey Kriegeskorte et al. Neuron 2008 Presented by Swaroop Guntupalli
Matching categorical object representation in the
inferior temporal cortex of man and monkey
Kriegeskorte et al.
Neuron 2008
Presented by
Swaroop Guntupalli
Why bother?
• Brain is a massive complex adaptive network (neurons are agents)
– 50-100 billion of these with 100 trillion synapses which change every second!
• So the problem is how to observe and understand its dynamics?
– Simple, just poke into it and record.
– Can’t do that on humans (since using torture on humans is only allowed for – Can’t do that on humans (since using torture on humans is only allowed for
homeland security but not for scientific understanding)
– fMRI (BOLD is alias of neural activity) in humans
– Do monkeys see the world like we do?
– How to compare activity of a few neurons to combined activity alias of a
cortical cube of side 2mm (which has tens of thousands of these units)
MA 076 Swaroop
Clever hack!
• Representational
dissimilarity matrix (RDM)
• It is based on stimuli , not
the modality of
measurement!
MA 076 Swaroop
So what?
• Response patterns in brain cluster according to same categorical structure in
monkey and man
• Stayed even if the regions that are specialized in processing faces and places are
removed! (Information representation is distributed)
• This categorization is absent in early visual processing areas (Visual cortex)
• Heirarchical category structure is inherent to IT• Heirarchical category structure is inherent to IT
• A low and intermediate level computer models did not produce such
categorization
• We and monkeys have a common code! (which is determined by yet another
complex network called gene regulatory network)
MA 076 Swaroop