Generalizing biological surround suppression based on center surround similarity via deep neural network models

Abstract

Sensory perception is dramatically influenced by the context. Models of contextual neural surround effects in vision have mostly accounted for Primary Visual Cortex (V1) data, via nonlinear computations such as divisive normalization. However, surround effects are not well understood within a hierarchy, for neurons with more complex stimulus selectivity beyond V1. We utilized feedforward deep convolutional neural networks and developed a gradient-based technique to visualize the most suppressive and excitatory surround. We found that deep neural networks exhibited a key signature of surround effects in V1, highlighting center stimuli that visually stand out from the surround and suppressing responses when the surround stimulus is similar to the center. We found that in some neurons, especially in late layers, when the center stimulus was altered, the most suppressive surround surprisingly can follow the change. Through the visualization approach, we generalized previous understanding of surround effects to more complex stimuli, in ways that have not been revealed in visual cortices. In contrast, the suppression based on center surround similarity was not observed in an untrained network. We identified further successes and mismatches of the feedforward CNNs to the biology. Our results provide a testable hypothesis of surround effects in higher visual cortices, and the visualization approach could be adopted in future biological experimental designs. Author summaryNeural responses and perception of a visual stimulus are influenced by the context, such as what spatially surrounds a given feature. Contextual surround effects have been extensively studied in the early visual cortex. But the brain processes visual inputs hierarchically, from simple features up to complex objects in higher visual areas. Contextual effects are not well understood for higher areas of cortex and for more complex stimuli. Utilizing artificial deep neural networks and a visualization technique we developed, we found that deep networks exhibited a key signature of surround effects in the early visual cortex, highlighting center stimuli that visually stand out from the surround and suppressing responses when the surround stimulus is similar to the center. We found in some neurons, especially in late layers, when the center stimulus was altered, the most suppressive surround could surprisingly follow. This is a generalization of known surround effects for more complex stimuli that has not been revealed in the visual cortex. Our findings relate to notions of efficient coding and salience perception, and emerged without incorporating specialized nonlinear computations typically used to explain contextual effects in the early cortex. Our visualization approach provides a new experimental paradigm and a testable hypothesis of surround effects for more complex stimuli in higher cortical areas; the visualization approach could be adopted in biological experimental designs.