DOE/Los Alamos National Laboratory

AI network simulations become unstable following continuous periods of unsupervised learning. When the networks were exposed to states that are analogous to human brain waves during sleep, the stability was restored.