Google Brain and DeepMind researchers attack strengthening learning proficiency

Reinforcement learning, which encourages AI to achieve goals through rewards or punishments, is a form of training that leads to gains in robotics, speech synthesis, and more. Unfortunately, he uses a lot of data that motivated research teams – one from Google Brain (one of Google’s AI research departments) and the other from Alphabet’s DeepMind – to create prototypes for more efficient ways to manage it. In some pre-printed articles, the researchers propose Adaptive Behavior Policy Sharing (ABPS), an algorithm that allows the sharing of experiences that are selected in an adaptable way by a pool of artificial intelligence agents and a structure – Universal Value Function Approximators (UVFA ) – which at the same time teaches exploration policy that focuses on the same AI, with different trade-offs between research and research.

Translate ยป