Timezone: »

 
Poster
Deep Learning Models of the Retinal Response to Natural Scenes
Lane McIntosh · Niru Maheswaranathan · Aran Nayebi · Surya Ganguli · Stephen Baccus

Mon Dec 05 09:00 AM -- 12:30 PM (PST) @ Area 5+6+7+8 #150
in Monday Posters »

A central challenge in sensory neuroscience is to understand neural computations and circuit mechanisms that underlie the encoding of ethologically relevant, natural stimuli. In multilayered neural circuits, nonlinear processes such as synaptic transmission and spiking dynamics present a significant obstacle to the creation of accurate computational models of responses to natural stimuli. Here we demonstrate that deep convolutional neural networks (CNNs) capture retinal responses to natural scenes nearly to within the variability of a cell's response, and are markedly more accurate than linear-nonlinear (LN) models and Generalized Linear Models (GLMs). Moreover, we find two additional surprising properties of CNNs: they are less susceptible to overfitting than their LN counterparts when trained on small amounts of data, and generalize better when tested on stimuli drawn from a different distribution (e.g. between natural scenes and white noise). An examination of the learned CNNs reveals several properties. First, a richer set of feature maps is necessary for predicting the responses to natural scenes compared to white noise. Second, temporally precise responses to slowly varying inputs originate from feedforward inhibition, similar to known retinal mechanisms. Third, the injection of latent noise sources in intermediate layers enables our model to capture the sub-Poisson spiking variability observed in retinal ganglion cells. Fourth, augmenting our CNNs with recurrent lateral connections enables them to capture contrast adaptation as an emergent property of accurately describing retinal responses to natural scenes. These methods can be readily generalized to other sensory modalities and stimulus ensembles. Overall, this work demonstrates that CNNs not only accurately capture sensory circuit responses to natural scenes, but also can yield information about the circuit's internal structure and function.

Keywords: (Other) Regression · (Other) Neuroscience · (Cognitive/Neuroscience) Neural Coding
[ PDF] 

Author Information

Lane McIntosh (Stanford University)
Niru Maheswaranathan (Stanford University)
Aran Nayebi (Stanford University)
Surya Ganguli (Stanford)
Stephen Baccus (Stanford University)

More from the Same Authors