ANNA is an attempt to make neural reinforcement learning efficient and scalable. ANNA is unique in that doesn't use backpropagation, and is instead based on a probabilistic graphical model.
Experimental Reinforcement Learning Architecture
ANNA is an attempt to make neural reinforcement learning efficient and scalable. ANNA is unique in that doesn't use backpropagation, and is instead based on a probabilistic graphical model.
Experimental Reinforcement Learning Architecture