Abstract
Divisive normalization (DN) is a prominent computational building block in the brain that has been proposed as a canonical cortical operation. Numerous experimental studies have verified its importance for capturing nonlinear neural response properties to simple, artificial stimuli, and computational studies suggest that DN is also an important component for processing natural stimuli. However, we lack quantitative models of DN that are directly informed by measurements of spiking responses in the brain and applicable to arbitrary stimuli. Here, we propose a DN model that is applicable to arbitrary input images. We test its ability to predict how neurons in macaque primary visual cortex (V1) respond to natural images, with a focus on nonlinear response properties within the classical receptive field. Our model consists of one layer of subunits followed by learned orientation-specific DN. It outperforms linear-nonlinear and wavelet-based feature representations and makes a significant step towards the performance of state-of-the-art convolutional neural network (CNN) models. Unlike deep CNNs, our compact DN model offers a direct interpretation of the nature of normalization. By inspecting the learned normalization pool of our model, we gained insights into a long-standing question about the tuning properties of DN that update the current textbook description: we found that within the receptive field oriented features were normalized preferentially by features with similar orientation rather than non-specifically as currently assumed.
Highlights
A crucial step towards understanding the visual system is to build models that predict neural responses to arbitrary stimuli with high accuracy [1]
We explored how our Divisive normalization (DN) model could be extended in a way that might enable it to capture surround interactions from outside of the receptive field (RF), our control experiments demonstrate that our results are very unlikely to include the extra-classical surround
Image-computable predictive model of divisive normalization following Eq (1), which is applicable to arbitrary images and whose parameters are learned by optimizing the accuracy of the model in predicting the spiking activity of a large number of neurons in response to natural images
Summary
A crucial step towards understanding the visual system is to build models that predict neural responses to arbitrary stimuli with high accuracy [1]. Fueled by advances in machine learning technology, recent studies have shown that multi-layer convolutional neural networks (CNNs) can significantly improve the prediction of neural responses to complex images and videos at several stages of the visual pathway, outperforming classical models [10,11,12,13,14,15,16,17]. The current state-of-the-art data-driven model of single-unit activity in monkey V1 is a three-layer CNN [14]. It is challenging to gain insights into V1 function from the features produced by deeper layers of such models. We do not have first principles explaining the kind of nonlinearities approximated in successive layers of CNNs, or if these nonlinearities can be described in a compact way in the first place
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
