AI-driven cholinergic theory enables rapid and robust cortex-wide learning

The cholinergic system has been associated with learning, but also with cognitive decline in dementia, aging and injury. Yet, to date, no computational models have been put forward to explain how the cholinergic system contributes to both learning and cognitive decline. Here we introduce a model that combines a recently proposed model of cortex-wide credit assignment with a cholinergic adaptive module based on adaptive deep learning rules. According to this model, the cholinergic system opens a cortex-wide gate for learning, but its end effect is controlled by local adaptive processes. Using a multi-class perceptual discrimination task we show that cholinergic adaptive learning leads to rapid cortex-wide learning when compared with non-adaptive models. Such cholinergic adaptive modulation results in a constant redistribution of learning across the cortex making task-encoding sparser. Consequently, we show that the network becomes more robust to perturbations such as simulated cell death. Moreover, we demonstrate that in order to obtain such rapid and robust learning, global mechanisms are not sufficient, suggesting the need for a tight cholinergic interaction with local cortical circuits. Overall, our work provides a novel theoretical framework for cellular-systems neuroscience with which to link cholinergic cortical modulation to health and disease.