A project of the U.S. Army has successfully developed a new framework for deep neural networks that makes artificial intelligence systems capable to learn new tasks better while forgetting less of what they have learned regarding previous tasks. Most of the AI systems today forget a part of the learning they have learned about the old tasks while learning about the new tasks. This limits the systems in learning new tasks. But with the new 'Learn to Grow' framework AI systems learn the new tasks while forgetting little about the old tasks. This makes the system a lifelong learner, just like a human being who continues to learn through the whole life, remembering the learning he had developed through various stages of his life.
The researchers at North Carolina State University also demonstrated that using the new 'Learn to Grow' framework the AI systems are better at performing previous tasks while learning new tasks. This phenomenon is called backward transfer.
"The Army needs to be prepared to fight anywhere in the world so it's intelligent systems also need to be prepared," said Dr. Mary Anne Fields, program manager for Intelligent Systems at Army Research Office, an element of U.S. Army Combat Capabilities Development Command's Army Research Lab. "We expect the Army's intelligent systems to continually acquire new skills as they conduct missions on battlefields around the world without forgetting skills that have already been trained. For instance, while conducting an urban operation, a wheeled robot may learn new navigation parameters for dense urban cities, but it still needs to operate efficiently in a previously encountered environment like a forest."
The new framework is expected to make the AI systems used by the Army stronger as the systems are ready to learn and acquire the new skill using the new framework while also remembering the old skills gained by it. This will make the system capable of handling the multiple tasks which will be effectively used by the Army in various combating operations.
The new framework outperformed the various previous learning systems which make it further more effective than any other framework. Also, the new framework is more accurate in performing older tasks than the previous methods, according to the researchers.
"Deep neural network AI systems are designed for learning narrow tasks," said Xilai Li, a co-lead author of the paper and a Ph.D. candidate at NC State. "As a result, one of several things can happen when learning new tasks, systems can forget old tasks when learning new ones, which is called catastrophic forgetting. Systems can forget some of the things they knew about old tasks, while not learning to do new ones as well. Or systems can fix old tasks in place while adding new tasks - which limits improvement and quickly leads to an AI system that is too large to operate efficiently. Continual learning, also called lifelong-learning or learning-to-learn, is trying to address the issue."
"We've run experiments using several datasets, and what we've found is that the more similar a new task is to previous tasks, the more overlap there is in terms of the existing layers that are kept to perform the new task," Li said. "What is more interesting is that with the optimized or "learned" topology - a network trained to perform new tasks forgets very little of what it needed to perform the older tasks, even if the older tasks were not similar."
"This Army investment extends the current state of the art machine learning techniques that will guide our Army Research Laboratory researchers as they develop robotic applications, such as intelligent maneuver and learning to recognize novel objects," Fields said. "This research brings AI a step closer to providing our warfighters with effective unmanned systems that can be deployed in the field."
Researchers have a great expectation from the Framework, "Learn to Grow: A Continual Structure Learning for Overcoming Catastrophic Forgetting," and the implementation of the framework in the AI systems is all that they are looking for now.