Discovering and Achieving Goals with World Models
Dec 6, 2021
Speakers
Russell Mendonca
Speaker · 0 followers
Oleh Rybkin
Speaker · 0 followers
Kostas Daniilidis
Speaker · 0 followers
Danijar Hafner
Speaker · 6 followers
Deepak Pathak
Speaker · 3 followers
About
How can an artificial agent learn to solve a wide range of tasks in a complex visual environment in the absence of external supervision? We decompose this question into two problems, global exploration of the environment and learning to reliably reach situations found during exploration. We introduce the Explore Achieve Network (ExaNet), a unified solution to these by learning a world model from the high-dimensional images and using it to train an explorer and an achiever policy from imagined trajectories. Unlike prior methods that explore by reaching previously visited states, our explorer plans to discover unseen surprising states through foresight, which are then used as diverse targets for the achiever. After the unsupervised phase, ExaNet solves tasks specified by goal images without any additional learning. We introduce a challenging benchmark spanning across four standard robotic manipulation and locomotion domains with a total of over 40 test tasks. Our agent substantially outperforms previous approaches to unsupervised goal reaching and achieves goals that require interacting with multiple objects in sequence. Finally, to demonstrate the scalability and generality of our approach, we train a single general agent across four distinct environments. For videos, see <https://sites.google.com/view/exanet/home>.How can an artificial agent learn to solve a wide range of tasks in a complex visual environment in the absence of external supervision? We decompose this question into two problems, global exploration of the environment and learning to reliably reach situations found during exploration. We introduce the Explore Achieve Network (ExaNet), a unified solution to these by learning a world model from the high-dimensional images and using it to train an explorer and an achiever policy from imagined t…
Organizer
NeurIPS 2021
Account · 1.9k followers
About NeurIPS 2021
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Role of Attachment Variables in Resilient Families
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Heterogeneous Multi-player Multi-armed Bandits: Closing the Gap and Generalization
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
DreamerPro: Reconstruction-Free Model-Based Reinforcement Learning with Prototypical Representations
Watch later
Fei Deng, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Near-Optimal Offline Reinforcement Learning via Double Variance Reduction
Watch later
Ming Yin, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Q-Pain: A Question Answering Dataset to Measure Social Bias in Pain Management
Watch later
Cécile Logé, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Task-Independent Causal State Abstraction
Watch later
Zizhao Wang, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%