Event Abstract

Category learning and decision making: a cortical circuit model

How does the brain recognize 'meaning' of sensory stimuli? Through experience we learn to group stimuli into arbitrary categories (such as 'animal' or 'car') according to certain shared characteristics. Recent neurophysiological studies have begun to investigate the neural mechanisms of categorization. Freedman and Assad (2006) recorded single-cell activity in monkeys classifying 360 of visual motion directions into two discrete categories. On each trial, the monkey had to indicate whether categories of two successively presented stimuli (sample and test) were the same (match) or different (nonmatch). Firing rates of neurons in the lateral intraparietal (LIP) area were found to encode the category membership of the currently visible stimulus and did not differentiate between physical features of the stimuli. In contrast, neurons in the middle temporal (MT) area were more involved in visual feature processing and did not carry explicit information about category. A similar categorization study (Freedman et al. 2003) revealed three neural populations in the prefrontal cortex (PFC): the first encoded category information, the second encoded match/nonmatch status of the test stimulus regardless of its category, and the third showed match/nonmatch effects that were limited to one of the categories.

The physiological data suggest that two main computational stages are involved in this task: (i) extract the category of the sample; and (ii) perform match vs. nonmatch comparison with the test. To identify general mechanisms underlying these basic computations, we propose a biophysically plausible network model of categorization and match vs. nonmatch comparison for motion directions. The model comprises two interconnected brain areas: a sensory (MT) and a cognitive-type area (LIP/PFC). Both areas are strongly recurrent circuits with dynamics governed by slow recurrent excitation and feedback inhibition. Importantly, heterogeneity of recurrent connections within the cognitive-type circuit leads to diversity of neural responses and a continuos spectrum of neural selectivities (e.g. mixed selectivity). We hypothesize that the bottom-up synapses from MT to LIP/PFC are plastic and implement the reward-dependent stochastic Hebbian learning rule to acquire new categories. The plastic synapses shape the direction-selective input from MT to produce category-tuned responses in LIP/PFC. Match vs. nonmatch decision in the model results from the winner-take-all competition between neural pools preferring matches and nonmatches. Along with the bottom-up input from MT, these neurons receive a weak modulatory input from the category selective neurons, which biases the competition and leads to enhancement/suppression of responses to the test stimulus, depending on whether it matches the sample category.

Our model accounts for the single-cell data and provides insight into general mechanisms of categorization and comparison of perceptual stimuli. The model predicts that (i) categories can be learned by synaptic plasticity in the input from sensory to cognitive-type areas; (ii) the mechanism of the match vs. nonmatch comparison is the biased winner-take-all competition.


1. D.J. Freedman and J. A. Assad, Nature 443, 85 (2006)

2. D.J. Freedman et al., J. Neurosci. 23, 5235 (2003)

Conference: Computational and systems neuroscience 2009, Salt Lake City, UT, United States, 26 Feb - 3 Mar, 2009.

Presentation Type: Poster Presentation

Topic: Poster Presentations

Citation: (2009). Category learning and decision making: a cortical circuit model. Front. Syst. Neurosci. Conference Abstract: Computational and systems neuroscience 2009. doi: 10.3389/conf.neuro.06.2009.03.055

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 30 Jan 2009; Published Online: 30 Jan 2009.