The most informative neural code accounts for population heterogeneity

Abstract

Perception is produced by reading out the representation of a sensory stimulus contained in the firing rates of a population of neurons. To examine experimentally how populations code information, a common approach is to decode a linearly-weighted sum of the neurons firing rates. This approach is popular because of its biological validity: weights in a computational decoder are analogous to synaptic strengths. For neurons recorded in vivo, weights are highly variable when derived through machine learning methods, but it is unclear what neuronal properties explain this variability, and how the variability affects decoding performance. To address this, we recorded from neurons in the middle temporal area (MT) of anesthetized marmosets (Callithrix jacchus) viewing stimuli comprising a sheet of dots that moved coherently in one of twelve different directions. We found that high gain and direction selectivity both predicted that a neuron would be weighted more highly in an optimised decoding model. Although learned weights differed markedly from weights chosen according to a priori rules based on a neurons tuning profile, decoding performance was only marginally better for the learned weights. In the models with a priori rules, selectivity is the best predictor of weighting, and defining weights according to a neurons preferred direction and selectivity improves decoding performance to very near the maximum level possible, as defined by the learned weights.\n\nNew & NoteworthyWe examined which aspects of a neurons tuning account for its contribution to sensory coding. Strongly direction-selective neurons were weighted most highly by machine learning algorithms trained to discriminate motion direction. Models with a priori defined decoding weights demonstrate that the learned weighting scheme causally improved direction representation by a neuronal population. Optimising decoders (using machine learning) lead to only marginally better performance than decoders based purely on a neurons preferred direction and selectivity.