A neuroplausible computational model of vision also exhibits asymmetry in developmental category learning


Computational models are increasingly used to explore possible mechanisms underlying infant capability in various tasks. Often, such models do not work on direct image data, but on hand-computed attributes of the images which are used as input in connectionist models. Such models are open to criticism since computing intermediate features may knowledge unavailable to the infant. Here we to explore the feasibility of the Serre-Poggio model which emulates the cortical ventral stream , and construct infant mental map using probabilistic models. In experiment 1, we consider asymmetry in visual category learning in early infancy (e.g. cats vs dogs), and show that surprisal for the novel category is higher when habituated on cat than on dog. Then we explore the role of face habituation and hierarchical category in categorization. These experiments suggest some mechanisms for the internal structures in infant learning, and also validate the S-P model for such tasks.

Back to Table of Contents