Model for computing with population-encoded variables explains neural correlations.

Decoding how the brain computes with variables, such as an object's speed, encoded in populations of neurons is essential for understanding perception, decision-making, and motor control. While experiments consistently reveal correlated firing rates within these populations, the origins of these correlations are still debated. Here, I propose a computational model that carries out operations on population-encoded variables by carefully configuring the connections between neurons through dendritic branches. This model predicts neural correlations as a byproduct of the computation, with correlation coefficients aligning with experimentally measured values in the literature. I also present a mathematical derivation of these coefficients, revealing their dependence on population size. To demonstrate the model's utility, I apply it to compute the prediction error between an observed object's speed and the predicted speed based on self-motion. Importantly, the framework generalizes to compute any functional relationship between population-encoded variables, serving as a building block for perception and behaviour, including tasks such as transforming an object's position into joint angles for grasping.