AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There a Model Base Learning for Atari?
After the dor mottos paper, there was another paper bout the model base learning for atari so and where they they deliterle call their aguism simple. The idea is basically, oc you have your agents collect data, train a generative model of the environment to predict the future, and then train your policy inside of that environment. So when that work came out in two thousand 19 at the time, that was then the state of the art for for sample efficiency for various atari games. In another line of work done by my colleague, dan jr afner, he started using these latent base rong models and combine them with planning agrithms