Homeostatic Adaptation of Optimal Population Codes under Metabolic Stress

Researchers have developed a novel AI framework that mimics how real neurons enter a 'low power mode' under metabolic stress. The model, based on mouse visual cortex data, shows neurons maintain function while expending less energy by flattening tuning curves and increasing noise. It introduces an energy budget tied to noise via biophysical simulation, creating an explainable mathematical framework that generalizes optimal population codes and connects ATP use directly to coding strategy.