Technical Program

Paper Detail

Paper: PS-2B.30
Session: Poster Session 2B
Location: Symphony/Overture
Session Time: Friday, September 7, 19:30 - 21:30
Presentation Time:Friday, September 7, 19:30 - 21:30
Presentation: Poster
Publication: 2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania
Paper Title: Linking image-by-image population dynamics in the macaque inferior temporal cortex to core object recognition behavior
Manuscript:  Click here to view manuscript
Authors: Kohitij Kar, Kailyn Schmidt, James DiCarlo, Massachusetts Institute of Technology, United States
Abstract: Primates can rapidly identify visual objects; an ability supported by the ventral visual stream. We have recently reported that object information emerges in the inferior temporal (IT) cortex with distinct image-dependent dynamics. However, the current most parsimonious model that accurately links IT neural activity to primates' core object recognition behavior involves learned weighted sums of IT firing rates, specifically integrating the IT evidence over a single, fixed time window (70 to 170 ms post image onset). Here we collected new data to test whether this baseline model could accurately predict image-level primate object confusion patterns and found that it could not fully do so. Therefore we built and tested a more biologically-plausible linking model that implements leaky IT evidence accumulation. This models accurately predicts the monkeys' image-by-image behavioral patterns tested on 45 binary object discrimination tasks. Furthermore, we discovered that the trial-by-trial behavior of this same model partly predicts the animal's trial-by-trial choices on ambiguous images. Taken together, these results argue that IT population dynamics are relevant to core object recognition behavior and we provide a new, improved model of the mechanistic linkage between IT and core object recognition behavior.