A More Effective Way To Train Machines For Uncertain, Real-world Situations

Researchers from MIT and Technion, the Israel Institute of Technology, have developed an algorithm that dynamically determines when a machine learning to complete a task should try to mimic its teacher or explore on its own through trial-and-error. This algorithm enabled simulated student machines to learn tasks faster and more effectively than other techniques. The combination of trial-and-error learning and imitation learning enabled students to learn tasks more effectively than methods that used only one type of learning, making it a useful tool for training machines for deployment in uncertain real-world environments.