Limits...
Combining computational modeling and neuroimaging to examine multiple category learning systems in the brain.

Nomura EM, Reber PJ - Brain Sci (2012)

Bottom Line: The PINNACLE (Parallel Interactive Neural Networks Active in Category Learning) model incorporates multiple categorization systems that compete to provide categorization judgments about visual stimuli.Incorporating competing systems requires inclusion of cognitive mechanisms associated with resolving this competition and creates a potential credit assignment problem in handling feedback.The hypothesized mechanisms make predictions about internal mental states that are not always reflected in choice behavior, but may be reflected in neural activity.

View Article: PubMed Central - PubMed

Affiliation: Helen Wills Neuroscience Institute, University of California, Berkeley, CA 94720, USA. eminomura@berkeley.edu.

ABSTRACT
Considerable evidence has argued in favor of multiple neural systems supporting human category learning, one based on conscious rule inference and one based on implicit information integration. However, there have been few attempts to study potential system interactions during category learning. The PINNACLE (Parallel Interactive Neural Networks Active in Category Learning) model incorporates multiple categorization systems that compete to provide categorization judgments about visual stimuli. Incorporating competing systems requires inclusion of cognitive mechanisms associated with resolving this competition and creates a potential credit assignment problem in handling feedback. The hypothesized mechanisms make predictions about internal mental states that are not always reflected in choice behavior, but may be reflected in neural activity. Two prior functional magnetic resonance imaging (fMRI) studies of category learning were re-analyzed using PINNACLE to identify neural correlates of internal cognitive states on each trial. These analyses identified additional brain regions supporting the two types of category learning, regions particularly active when the systems are hypothesized to be in maximal competition, and found evidence of covert learning activity in the "off system" (the category learning system not currently driving behavior). These results suggest that PINNACLE provides a plausible framework for how competing multiple category learning systems are organized in the brain and shows how computational modeling approaches and fMRI can be used synergistically to gain access to cognitive processes that support complex decision-making machinery.

No MeSH data available.


RB (A) and II stimuli (B). Each point represents a distinct Gabor patch (sine-wave) stimulus defined by orientation (tilt) and frequency (thickness of lines). In both stimulus sets, there are two categories (red and blue points). RB categories are defined by a vertical boundary (only frequency is relevant for categorization) whereas II categories are defined by a diagonal boundary (both orientation and frequency are relevant). In both RB and II stimuli there are examples of a stimulus from each category. (C) Schematic of a single trial. A fixation point is followed by the to-be-categorized-stimulus (either RB or II depending on the subject), then a short visual mask that is followed by the feedback. The subject responded “category A” or “category B” during the 2 s the stimulus was on the screen using hand-held buttons. The length of the inter-trial interval (ITI) was pseudorandom and based on between zero and five 4-s “fixation-only” trial periods arranged to maximize the separability of the measured hemodynamic response to stimulus trials.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4061791&req=5

brainsci-02-00176-f001: RB (A) and II stimuli (B). Each point represents a distinct Gabor patch (sine-wave) stimulus defined by orientation (tilt) and frequency (thickness of lines). In both stimulus sets, there are two categories (red and blue points). RB categories are defined by a vertical boundary (only frequency is relevant for categorization) whereas II categories are defined by a diagonal boundary (both orientation and frequency are relevant). In both RB and II stimuli there are examples of a stimulus from each category. (C) Schematic of a single trial. A fixation point is followed by the to-be-categorized-stimulus (either RB or II depending on the subject), then a short visual mask that is followed by the feedback. The subject responded “category A” or “category B” during the 2 s the stimulus was on the screen using hand-held buttons. The length of the inter-trial interval (ITI) was pseudorandom and based on between zero and five 4-s “fixation-only” trial periods arranged to maximize the separability of the measured hemodynamic response to stimulus trials.

Mentions: For visual categories, category structure can be efficiently described as partitioning perceptual space and assigning category labels (or motor responses) to regions that encompass a collection of similar stimuli. A mathematical description of this approach is the core of the decision-bound theory (DBT) of category learning first described by Ashby and Gott [14]. Under DBT, when presented with a to-be-categorized stimulus, participants determine in what region the stimulus has fallen and produce the associated response. Feedback about the accuracy of the judgment is used to adjust the boundaries to improve future categorization performance. In this approach, learning the categories is equivalent to identifying the optimal decision-boundary that separates the categories in the perceptual space. In most studies of RB and II category learning, stimuli are constructed in a 2-dimensional space with continuous dimensions. An example of 2-dimensional stimulus space and the corresponding category bounds is shown in Figure 1.


Combining computational modeling and neuroimaging to examine multiple category learning systems in the brain.

Nomura EM, Reber PJ - Brain Sci (2012)

RB (A) and II stimuli (B). Each point represents a distinct Gabor patch (sine-wave) stimulus defined by orientation (tilt) and frequency (thickness of lines). In both stimulus sets, there are two categories (red and blue points). RB categories are defined by a vertical boundary (only frequency is relevant for categorization) whereas II categories are defined by a diagonal boundary (both orientation and frequency are relevant). In both RB and II stimuli there are examples of a stimulus from each category. (C) Schematic of a single trial. A fixation point is followed by the to-be-categorized-stimulus (either RB or II depending on the subject), then a short visual mask that is followed by the feedback. The subject responded “category A” or “category B” during the 2 s the stimulus was on the screen using hand-held buttons. The length of the inter-trial interval (ITI) was pseudorandom and based on between zero and five 4-s “fixation-only” trial periods arranged to maximize the separability of the measured hemodynamic response to stimulus trials.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4061791&req=5

brainsci-02-00176-f001: RB (A) and II stimuli (B). Each point represents a distinct Gabor patch (sine-wave) stimulus defined by orientation (tilt) and frequency (thickness of lines). In both stimulus sets, there are two categories (red and blue points). RB categories are defined by a vertical boundary (only frequency is relevant for categorization) whereas II categories are defined by a diagonal boundary (both orientation and frequency are relevant). In both RB and II stimuli there are examples of a stimulus from each category. (C) Schematic of a single trial. A fixation point is followed by the to-be-categorized-stimulus (either RB or II depending on the subject), then a short visual mask that is followed by the feedback. The subject responded “category A” or “category B” during the 2 s the stimulus was on the screen using hand-held buttons. The length of the inter-trial interval (ITI) was pseudorandom and based on between zero and five 4-s “fixation-only” trial periods arranged to maximize the separability of the measured hemodynamic response to stimulus trials.
Mentions: For visual categories, category structure can be efficiently described as partitioning perceptual space and assigning category labels (or motor responses) to regions that encompass a collection of similar stimuli. A mathematical description of this approach is the core of the decision-bound theory (DBT) of category learning first described by Ashby and Gott [14]. Under DBT, when presented with a to-be-categorized stimulus, participants determine in what region the stimulus has fallen and produce the associated response. Feedback about the accuracy of the judgment is used to adjust the boundaries to improve future categorization performance. In this approach, learning the categories is equivalent to identifying the optimal decision-boundary that separates the categories in the perceptual space. In most studies of RB and II category learning, stimuli are constructed in a 2-dimensional space with continuous dimensions. An example of 2-dimensional stimulus space and the corresponding category bounds is shown in Figure 1.

Bottom Line: The PINNACLE (Parallel Interactive Neural Networks Active in Category Learning) model incorporates multiple categorization systems that compete to provide categorization judgments about visual stimuli.Incorporating competing systems requires inclusion of cognitive mechanisms associated with resolving this competition and creates a potential credit assignment problem in handling feedback.The hypothesized mechanisms make predictions about internal mental states that are not always reflected in choice behavior, but may be reflected in neural activity.

View Article: PubMed Central - PubMed

Affiliation: Helen Wills Neuroscience Institute, University of California, Berkeley, CA 94720, USA. eminomura@berkeley.edu.

ABSTRACT
Considerable evidence has argued in favor of multiple neural systems supporting human category learning, one based on conscious rule inference and one based on implicit information integration. However, there have been few attempts to study potential system interactions during category learning. The PINNACLE (Parallel Interactive Neural Networks Active in Category Learning) model incorporates multiple categorization systems that compete to provide categorization judgments about visual stimuli. Incorporating competing systems requires inclusion of cognitive mechanisms associated with resolving this competition and creates a potential credit assignment problem in handling feedback. The hypothesized mechanisms make predictions about internal mental states that are not always reflected in choice behavior, but may be reflected in neural activity. Two prior functional magnetic resonance imaging (fMRI) studies of category learning were re-analyzed using PINNACLE to identify neural correlates of internal cognitive states on each trial. These analyses identified additional brain regions supporting the two types of category learning, regions particularly active when the systems are hypothesized to be in maximal competition, and found evidence of covert learning activity in the "off system" (the category learning system not currently driving behavior). These results suggest that PINNACLE provides a plausible framework for how competing multiple category learning systems are organized in the brain and shows how computational modeling approaches and fMRI can be used synergistically to gain access to cognitive processes that support complex decision-making machinery.

No MeSH data available.