2017/06/07 by Joshua C. Peterson, Peterson, Joshua C., Joshua T. Abbott +3
Computer Science · Neuroscience · Psychology · #Child and Animal Learning Development #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Face Recognition and Perception
paper · pdf · doi:10.48550/arxiv.1706.02417
openalex publication_date 2017/06/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Decades of psychological research have been aimed at modeling how people\nlearn features and categories. The empirical validation of these theories is\noften based on artificial stimuli with simple representations. Recently, deep\nneural networks have reached or surpassed human accuracy on tasks such as\nidentifying objects in natural images. These networks learn representations of\nreal-world stimuli that can potentially be leveraged to capture psychological\nrepresentations. We find that state-of-the-art object classification networks\nprovide surprisingly accurate predictions of human similarity judgments for\nnatural images, but fail to capture some of the structure represented by\npeople. We show that a simple transformation that corrects these discrepancies\ncan be obtained through convex optimization. We use the resulting\nrepresentations to predict the difficulty of learning novel categories of\nnatural images. Our results extend the scope of psychological experiments and\ncomputational modeling by enabling tractable use of large natural stimulus\nsets.\n