August 2023
Volume 23, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   August 2023
A neural network model of category-learning induced transfer of visual perceptual learning
Author Affiliations
  • Luke Rosedahl
    Brown University
  • Thomas Serre
    Brown University
  • Takeo Watanabe
    Brown University
Journal of Vision August 2023, Vol.23, 5378. doi:https://doi.org/10.1167/jov.23.9.5378
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Luke Rosedahl, Thomas Serre, Takeo Watanabe; A neural network model of category-learning induced transfer of visual perceptual learning. Journal of Vision 2023;23(9):5378. https://doi.org/10.1167/jov.23.9.5378.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Visual Perceptual Learning (VPL; often defined as a long-term performance increase resulting from visual experience) is highly specific to trained features. Previous work found that performing category learning before VPL causes VPL to transfer to stimuli from the same category as the trained stimulus (Category-Learning Induced Transfer of VPL or CIT-VPL, Wang et al, 2018, Current Biology). However, the mechanism of transfer is unknown. Based on work showing that Feature-Based Attention (FBA) during category learning can increase within-category stimulus similarity (Brouwer and Heeger, 2013), here we postulate that CIT-VPL occurs through FBA. We test this hypothesis utilizing two category structures: Rule-Based (RB) and Information-Integration (II). In RB structures, the optimal strategy involves making binary decisions along feature dimensions and performance is increased if FBA is targeted to specific feature values. In II structures, information from multiple feature dimensions must be combined before a decision can be made and performance is decreased if FBA is targeted to specific feature values. The theory therefore predicts that RB structures will cause greater transfer of VPL than II structures. Subjects (n=10) were divided evenly between RB and II conditions and underwent category learning followed by five days of VPL training. VPL for the trained stimulus and a transfer stimulus from each category was measured using pre- and post-testing. The RB condition induced transfer for the stimulus from the same category as the trained stimulus but not for the opposing category stimulus, while the II condition induced no transfer. We then implement a neural network model that learns to apply feature-specific feedback (gain) modulation during category learning. We demonstrate that feedback connections enable the network to show the same transfer patterns as humans. Overall, this work provides computational and behavioral evidence for feature-based attention being the mechanism for category-learning induced transfer of VPL.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×