01 02 03 04 05
06 07 08 09 10 11 12
13 14 15 16 17 18 19
20 21 22 23 24 25 26
27 28 29 30 31    
Add An Event

CCN Brown Bag Series

Wednesday, November 28, 2018,

Event Image
  • Location: Wilson Hall • 111 21St Ave S • Nashville, TN 37240
  • Room: 115

Craig Sanders, PhD

Department of Psychology (Palmeri Lab)

Vanderbilt University

"Using deep learning to extract psychological representations of natural stimuli"

 Tests of formal models of human categorization have traditionally been restricted to artificial categories because deriving psychological representations for large numbers of natural stimuli has been an intractable task. In this talk, I show that deep learning may be used to solve this problem. I train convolutional neural networks (CNNs) to produce the multidimensional scaling (MDS) coordinates of images of rocks. I then show that not only are the CNNs able to predict the MDS coordinates of a held-out test set of rocks, but that the CNN-derived representations can be used in combination with  a formal psychological model to predict human categorization behavior on a completely new set of rocks. Finally, I demonstrate the flexibility of this approach by showing that the CNNs can also be trained to produce dimensions that are not in the original MDS space but are nonetheless psychologically-meaningful and important for making categorization decisions.