Li and Atick's theory of efficient binocular coding: A tutorial and mini-review. (December 2022)

• Record Type: Journal Article
• Authors: May, Keith; Zhaoping, Li
• Is Part Of: Vision research. Volume 201(2022)
• Publication Date: 2022-12
• Subjects: Binocular vision -- Efficient coding -- Theoretical neuroscience; Vision -- Periodicals; 573.88
• DOI: 10.1016/j.visres.2021.08.005
• Languages: English
• ISSNs: 0042-6989

Abstract

Abstract: Li and Atick (Network: Computation in Neural Systems 5 (1994) 157–174) presented a theory of efficient binocular encoding that explains a number of experimental findings. A binocular neuron is conventionally described in terms of two channels: the left and right eyes. Li and Atick's theory instead describes the neuron in terms of two alternative channels: the binocular sum and difference. The advantage of the latter description is that, unlike the left and right eye channels, the summation and differencing channels are usually uncorrelated; this means that each channel can be optimised independently of the other. The theory shows how to derive optimal receptive fields for the binocular summation and differencing channels; from these, it is easy to derive the neuron's optimal left and right eye receptive fields. The functional reality of the summation and differencing channels is demonstrated by a series of adaptation studies that confirm some counterintuitive predictions of the theory. Here we provide an accessible account of the theory, and review the evidence supporting it.