Sign In

Communications of the ACM

ACM TechNews

AI Uses Artificial Sleep to Learn New Task without Forgetting the Last

View as: Print Mobile App Share:
The researchers simulated sleep in the neural network by activating the networks artificial neurons in a noisy pattern.

Experiments demonstrated the importance of having “rapidly alternating sessions of training and sleep” while the AI was learning a second task.

Credit: Los Alamos National Laboratory

Scientists at the University of California, San Diego (UCSD) and the Czech Republic's Czech Academy of Sciences taught an artificial intelligence (AI) to learn a second distinct task without overwriting connections learned from a first task, through the use of simulated sleep.

UCSD's Erik Delanois said it was critical to "have rapidly alternating sessions of training and sleep" while the AI was learning the second task, which consolidated links from the first task that would have otherwise been forgotten.

"Such a network will have the ability to combine consecutively learned knowledge in smart ways, and apply this learning to novel situations—just like animals and humans do," said the University of Massachusetts Amherst's Hava Siegelmann.

From New Scientist
View Full Article


Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account