Do Neural Networks Need Sleep?

Study prescribes sleep-like brain waves as a remedy for instability in neuromorphic simulations

Researchers at Los Alamos National Laboratory have discovered that neural networks designed to behave like human brains might need occasional doses of sleep-like brain waves in order to function optimally. Los Alamos computer scientist Yijing Watkins and her team discovered that simulations run on neuromorphic processors that are trained “in a manner analogous to how humans and other biological systems learn from their environment...” experience instability after continuous periods of unsupervised learning.

According to the press release, “When they exposed the networks to states that are analogous to the waves that living brains experience during sleep, stability was restored. ‘It was as though we were giving the neural networks the equivalent of a good night’s rest,’ said Watkins.