CNN explains tuning properties of anterior, but not middle, face-processing areas in macaque IT

Abstract

Recent computational studies have emphasized layer-wise quantitative similarity between convolutional neural networks (CNNs) and the primate visual ventral stream. However, whether such similarity holds for the face-selective areas, a subsystem of the higher visual cortex, is not clear. Here, we extensively investigate whether CNNs exhibit tuning properties as previously observed in different macaque face areas. While simulating four past experiments on a variety of CNN models, we sought for the model layer that quantitatively matches the multiple tuning properties of each face area. Our results show that higher model layers explain reasonably well the properties of anterior areas, while no layer simultaneously explains the properties of middle areas, consistently across the model variation. Thus, the CNNs may have some similarity with the primate face-processing system in the near-goal representation, but differ in the intermediate computational process, thus requiring a more comprehensive model for understanding the entire system.Recently, the neuroscience community has witnessed the rise of the deep convolution neural network (CNN) 1 , a family of feedforward artificial neural networks, in computational modeling of the primate visual system. CNN models trained for behavioral goals have exhibited remarkable similarity to ventral visual areas in terms of stimulus-response relationship despite that the network itself was not directly optimized to fit neural data. For

Explaining face representation in the primate brain using different computational models

“…4 At the same time, a number of groups have found that activity in deep layers of convolutional neural networks can explain significant variance of neural responses in ventral temporal cortex. 26,[29][30][31][32] Here, we extend these earlier results by comparing the efficacy of a large number of different computational models of face representation to account for neural activity in face patch AM. We were especially interested in how the 2D morphable model, a simple and explicit graphical model, would compare to VGG-face, a black box deep neural network dedicated to face recognition containing hundreds of thousands of parameters and trained on nearly a million (982,803) facial images.…”