Explore the environment by learning a random neural network скачать в хорошем качестве

Explore the environment by learning a random neural network

Environments with extremely sparse reward are a incredibly tough to solve for the current reinforcement learning algorithms. In this Video, you'll learn how Random Network Distillation from the paper "Exploration By Random Network Distillation" works and how an agent can explore an environment without any supervision signal, by only using an intrinsic reward. Enjoy the video! Papers cited in the video: Exploration By Random Network Distillation https://arxiv.org/abs/1810.12894 To learn more about exploration in Deep Reinforcement Learning, take a look at this article by Lilian Weng: Exploration Strategies in Deep Reinforcement Learning (https://lilianweng.github.io/lil-log/...) Paper abstract ### We introduce an exploration bonus for deep reinforcement learning methods that is easy to implement and adds minimal overhead to the computation performed. The bonus is the error of a neural network predicting features of the observations given by a fixed randomly initialized neural network. We also introduce a method to flexibly combine intrinsic and extrinsic rewards. We find that the random network distillation (RND) bonus combined with this increased flexibility enables significant progress on several hard exploration Atari games. In particular we establish state of the art performance on Montezuma's Revenge, a game famously difficult for deep reinforcement learning methods. To the best of our knowledge, this is the first method that achieves better than average human performance on this game without using demonstrations or having access to the underlying state of the game, and occasionally completes the first level. Images credit: Berkeley CS188 #reinforcementlearning #randomnetworkdistillation #deeplearning #exploration