Psychological and Brain Sciences
On any given day, we receive a stream of visual information that is sampled from the environment in the form of retinal images. Exactly how the early visual system transforms this onslaught of retinal images into unique neural representations is a long-standing question in systems neuroscience. In the last three decades, a wide range of important insights into visual processing have been made by considering how the statistical properties of natural scenes relate to the response properties of the visual system. This work has enabled a better understanding of receptive field response dynamics and optimized population coding-strategies for natural scenes, and has begun to establish what sources of statistical structure are encoded in the face of redundancy across different natural scenes. Despite these advancements in our understanding of how the visual system codes for natural scenes, we still lack a clear understanding of how the early visual code transforms and ultimately organizes scenes from different visual environments.
One approach to conceptualize this problem is to consider the geometric organization of images according to the responses of all neurons within the early visual system. What such a state-space conceptualization essentially does is consider where each image would be plotted in an n-dimensional geometric space, with each axis of that space corresponding to a given neuron in the early visual system. The location of any given image along each neuron’s axis could be determined, for example, by the firing rate of each neuron. This project therefore adopts the geometric approach and aims to 1) examine the early encoding of natural scene images by the visual system through state-space geometry as indexed by macroscale (EEG) neural signals, and 2) to describe the transformation of the natural image state-space into a neural response space defined by EEG.
Must be intrinsically motivated and possess a genuine passion for learning.