Multiple homeostatic plasticity mechanisms are thought to be critical for the prevention of excessively high or aberrantly low neural activity in the adult cortex. In L2/3 of adult mouse visual cortex the interplay between disinhibition and local functional interactions may support homeostatic recovery following visual deprivation. Despite blanket disinhibition only a subset of L2/3 excitatory neurons are observed to exhibit homeostatic recovery. Recovering neurons tend to be correlated with each other, forming functional networks prior to deprivation. How homeostatic recovery occurs in this way is therefore unclear, particularly in conditions of global disinhibition. Here, we employ a computational modelling approach to investigate the origin of diverse homeostatic responses in the cortex. This model finds network size to be a critical determinant of the diverse homeostatic activity profiles observed following visual deprivation, as neurons which belong to larger networks exhibit a stronger homeostatic response. Our simulations provide mechanistic insights into the emergence of diverse homeostatic responses, and predict that neurons with a high proportion of enduring functional associations will exhibit the strongest homeostatic recovery. We test and confirm these predictions experimentally.