Dream to Control: Learning Behaviors by Latent Imagination
Dreamer is a new RL agent by DeepMind that learns a continuous control task through forward-imagination in latent space.
https://arxiv.org/abs/1912.01603
Videos: https://dreamrl.github.io/
Abstract:
Learned world models summarize an agent's experience to facilitate learning complex behaviors. While learning world models from high-dimensional sensory inputs is becoming feasible through deep learning, there are many potential ways for deriving behaviors from them. We present Dreamer, a reinforcement learning agent that solves long-horizon tasks from images purely by latent imagination. We efficiently learn behaviors by propagating analytic gradients of learned state values back through trajectories imagined in the compact state space of a learned world model. On 20 challenging visual control tasks, Dreamer exceeds existing approaches in data-efficiency, computation time, and final performance.
Authors: Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi
Links:
YouTube: https://www.youtube.com/c/yannickilcher
Twitter: https://twitter.com/ykilcher
BitChute: https://www.bitchute.com/channel/yannic-kilcher
Minds: https://www.minds.com/ykilcher
Видео Dream to Control: Learning Behaviors by Latent Imagination канала Yannic Kilcher
https://arxiv.org/abs/1912.01603
Videos: https://dreamrl.github.io/
Abstract:
Learned world models summarize an agent's experience to facilitate learning complex behaviors. While learning world models from high-dimensional sensory inputs is becoming feasible through deep learning, there are many potential ways for deriving behaviors from them. We present Dreamer, a reinforcement learning agent that solves long-horizon tasks from images purely by latent imagination. We efficiently learn behaviors by propagating analytic gradients of learned state values back through trajectories imagined in the compact state space of a learned world model. On 20 challenging visual control tasks, Dreamer exceeds existing approaches in data-efficiency, computation time, and final performance.
Authors: Danijar Hafner, Timothy Lillicrap, Jimmy Ba, Mohammad Norouzi
Links:
YouTube: https://www.youtube.com/c/yannickilcher
Twitter: https://twitter.com/ykilcher
BitChute: https://www.bitchute.com/channel/yannic-kilcher
Minds: https://www.minds.com/ykilcher
Видео Dream to Control: Learning Behaviors by Latent Imagination канала Yannic Kilcher
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
![Agent57: Outperforming the Atari Human Benchmark](https://i.ytimg.com/vi/U3zmekzQ8WQ/default.jpg)
![Dream to Control: Learning Behaviors by Latent Imagination](https://i.ytimg.com/vi/BDxRNnhPTlU/default.jpg)
![MuZero: Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model](https://i.ytimg.com/vi/We20YSAJZSE/default.jpg)
![How to escape education's death valley | Sir Ken Robinson](https://i.ytimg.com/vi/wX78iKhInsc/default.jpg)
![The Super Mario Effect - Tricking Your Brain into Learning More | Mark Rober | TEDxPenn](https://i.ytimg.com/vi/9vJRopau0g0/default.jpg)
![NFNets: High-Performance Large-Scale Image Recognition Without Normalization (ML Paper Explained)](https://i.ytimg.com/vi/rNkHjZtH0RQ/default.jpg)
![Space Elevator – Science Fiction or the Future of Mankind?](https://i.ytimg.com/vi/qPQQwqGWktE/default.jpg)
![Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://i.ytimg.com/vi/iAR8LkkMMIM/default.jpg)
![Charles Blundell - Agent57: Outperforming the Atari Human Benchmark](https://i.ytimg.com/vi/VQEg8aSpXcU/default.jpg)
![GLOM: How to represent part-whole hierarchies in a neural network (Geoff Hinton's Paper Explained)](https://i.ytimg.com/vi/cllFzkvrYmE/default.jpg)
![The Visual Task Adaptation Benchmark](https://i.ytimg.com/vi/ctCv_NRpqvM/default.jpg)
![Are Space Elevators Possible?](https://i.ytimg.com/vi/Xa_xteu_Mts/default.jpg)
![Dreamer v2: Mastering Atari with Discrete World Models (Machine Learning Research Paper Explained)](https://i.ytimg.com/vi/o75ybZ-6Uu8/default.jpg)
![Axial Attention & MetNet: A Neural Weather Model for Precipitation Forecasting](https://i.ytimg.com/vi/lmAj0SU_bW0/default.jpg)
![Imitation Learning: Reinforcement Learning For The Real World - Dr. Byron Galbraith](https://i.ytimg.com/vi/K_c2ko4AKgg/default.jpg)
![PlaNet ICML 2019 Talk](https://i.ytimg.com/vi/novKgDBa_2Y/default.jpg)
![Google’s PlaNet AI Learns Planning from Pixels](https://i.ytimg.com/vi/C7Dmu0GtrSw/default.jpg)
![OpenAI CLIP: ConnectingText and Images (Paper Explained)](https://i.ytimg.com/vi/T9XSU0pKX2E/default.jpg)
![Melanie Zeilinger: "Learning-based Model Predictive Control - Towards Safe Learning in Control"](https://i.ytimg.com/vi/nO8r8XKlPgA/default.jpg)
![Reformer: The Efficient Transformer](https://i.ytimg.com/vi/i4H0kjxrias/default.jpg)