Scaling data-driven robotics with reward sketching and batch reinforcement learning
We present a framework for data-driven robotics that makes use of a large dataset of
recorded robot experience and scales to several tasks using learned reward functions. We …
recorded robot experience and scales to several tasks using learned reward functions. We …
Asking easy questions: A user-friendly approach to active reward learning
Robots can learn the right reward function by querying a human expert. Existing approaches
attempt to choose questions where the robot is most uncertain about the human's response; …
attempt to choose questions where the robot is most uncertain about the human's response; …
Machine teaching for inverse reinforcement learning: Algorithms and applications
Inverse reinforcement learning (IRL) infers a reward function from demonstrations, allowing
for policy improvement and generalization. However, despite much recent interest in IRL …
for policy improvement and generalization. However, despite much recent interest in IRL …
Interactive teaching algorithms for inverse reinforcement learning
We study the problem of inverse reinforcement learning (IRL) with the added twist that the
learner is assisted by a helpful teacher. More formally, we tackle the following algorithmic …
learner is assisted by a helpful teacher. More formally, we tackle the following algorithmic …
Incomplete contracting and AI alignment
D Hadfield-Menell, GK Hadfield - Proceedings of the 2019 AAAI/ACM …, 2019 - dl.acm.org
We suggest that the analysis of incomplete contracting developed by law and economics
researchers can provide a useful framework for understanding the AI alignment problem and …
researchers can provide a useful framework for understanding the AI alignment problem and …
Preferences implicit in the state of the world
R Shah, D Krasheninnikov, J Alexander… - arXiv preprint arXiv …, 2019 - arxiv.org
Reinforcement learning (RL) agents optimize only the features specified in a reward function
and are indifferent to anything left out inadvertently. This means that we must not only …
and are indifferent to anything left out inadvertently. This means that we must not only …
Batch active learning using determinantal point processes
Data collection and labeling is one of the main challenges in employing machine learning
algorithms in a variety of real-world applications with limited data. While active learning …
algorithms in a variety of real-world applications with limited data. While active learning …
Active learning of reward dynamics from hierarchical queries
Enabling robots to act according to human preferences across diverse environments is a
crucial task, extensively studied by both roboticists and machine learning researchers. To …
crucial task, extensively studied by both roboticists and machine learning researchers. To …
The green choice: Learning and influencing human decisions on shared roads
Autonomous vehicles have the potential to increase the capacity of roads via platooning,
even when human drivers and autonomous vehicles share roads. However, when users of a …
even when human drivers and autonomous vehicles share roads. However, when users of a …
Verifying robustness of human-aware autonomous cars
As human-robot systems make their ways into our every day life, safety has become a core
concern of the learning algorithms used by such systems. Examples include semi …
concern of the learning algorithms used by such systems. Examples include semi …