Towards a deep learning model of retina: Retinal neural encoding of color flash patterns

Abstract

The retina is the first stage of visual neural information coding on the visual system, and several challenges remain on its functioning. Overcoming these challenges would suppose both a step further in the general understanding of the biological neural systems and a potential way to enhance millions of people’s lives that suffer from visual degeneration or impairment. In this work, a data-driven deep learning approach is applied to learn the behavior of mice’s retinal ganglion cells in response to light, as a step towards the development of a system able to mimic a real retina in terms of neural coding of visual stimuli.