Ultra-fast Object Recognition from Few Spikes

Understanding the complex brain computations leading to object recognition requires quantitatively characterizing the information represented in inferior temporal cortex (IT), the highest stage of the primate visual stream. A read-out technique based on a trainable classifier is used to characterize...

Full description

Bibliographic Details
Main Authors: Hung, Chou, Kreiman, Gabriel, Poggio, Tomaso, DiCarlo, James J.
Language:en_US
Published: 2005
Subjects:
Online Access:http://hdl.handle.net/1721.1/30556
_version_ 1826217962122510336
author Hung, Chou
Kreiman, Gabriel
Poggio, Tomaso
DiCarlo, James J.
author_facet Hung, Chou
Kreiman, Gabriel
Poggio, Tomaso
DiCarlo, James J.
author_sort Hung, Chou
collection MIT
description Understanding the complex brain computations leading to object recognition requires quantitatively characterizing the information represented in inferior temporal cortex (IT), the highest stage of the primate visual stream. A read-out technique based on a trainable classifier is used to characterize the neural coding of selectivity and invariance at the population level. The activity of very small populations of independently recorded IT neurons (~100 randomly selected cells) over very short time intervals (as small as 12.5 ms) contains surprisingly accurate and robust information about both object ‘identity’ and ‘category’, which is furthermore highly invariant to object position and scale. Significantly, selectivity and invariance are present even for novel objects, indicating that these properties arise from the intrinsic circuitry and do not require object-specific learning. Within the limits of the technique, there is no detectable difference in the latency or temporal resolution of the IT information supporting so-called ‘categorization’ (a.k. basic level) and ‘identification’ (a.k. subordinate level) tasks. Furthermore, where information, in particular information about stimulus location and scale, can also be read-out from the same small population of IT neurons. These results show how it is possible to decode invariant object information rapidly, accurately and robustly from a small population in IT and provide insights into the nature of the neural code for different kinds of object-related information.
first_indexed 2024-09-23T17:11:48Z
id mit-1721.1/30556
institution Massachusetts Institute of Technology
language en_US
last_indexed 2024-09-23T17:11:48Z
publishDate 2005
record_format dspace
spelling mit-1721.1/305562019-04-12T13:39:28Z Ultra-fast Object Recognition from Few Spikes Hung, Chou Kreiman, Gabriel Poggio, Tomaso DiCarlo, James J. AI object recognition neural coding inferior temporal cortex Understanding the complex brain computations leading to object recognition requires quantitatively characterizing the information represented in inferior temporal cortex (IT), the highest stage of the primate visual stream. A read-out technique based on a trainable classifier is used to characterize the neural coding of selectivity and invariance at the population level. The activity of very small populations of independently recorded IT neurons (~100 randomly selected cells) over very short time intervals (as small as 12.5 ms) contains surprisingly accurate and robust information about both object ‘identity’ and ‘category’, which is furthermore highly invariant to object position and scale. Significantly, selectivity and invariance are present even for novel objects, indicating that these properties arise from the intrinsic circuitry and do not require object-specific learning. Within the limits of the technique, there is no detectable difference in the latency or temporal resolution of the IT information supporting so-called ‘categorization’ (a.k. basic level) and ‘identification’ (a.k. subordinate level) tasks. Furthermore, where information, in particular information about stimulus location and scale, can also be read-out from the same small population of IT neurons. These results show how it is possible to decode invariant object information rapidly, accurately and robustly from a small population in IT and provide insights into the nature of the neural code for different kinds of object-related information. 2005-12-22T02:33:06Z 2005-12-22T02:33:06Z 2005-07-06 MIT-CSAIL-TR-2005-045 AIM-2005-022 CBCL-253 http://hdl.handle.net/1721.1/30556 en_US Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory 30 p. 77109103 bytes 12556007 bytes application/postscript application/pdf application/postscript application/pdf
spellingShingle AI
object recognition
neural coding
inferior temporal cortex
Hung, Chou
Kreiman, Gabriel
Poggio, Tomaso
DiCarlo, James J.
Ultra-fast Object Recognition from Few Spikes
title Ultra-fast Object Recognition from Few Spikes
title_full Ultra-fast Object Recognition from Few Spikes
title_fullStr Ultra-fast Object Recognition from Few Spikes
title_full_unstemmed Ultra-fast Object Recognition from Few Spikes
title_short Ultra-fast Object Recognition from Few Spikes
title_sort ultra fast object recognition from few spikes
topic AI
object recognition
neural coding
inferior temporal cortex
url http://hdl.handle.net/1721.1/30556
work_keys_str_mv AT hungchou ultrafastobjectrecognitionfromfewspikes
AT kreimangabriel ultrafastobjectrecognitionfromfewspikes
AT poggiotomaso ultrafastobjectrecognitionfromfewspikes
AT dicarlojamesj ultrafastobjectrecognitionfromfewspikes