Preference-Based Policy Iteration: Leveraging Preference Learning for Reinforcement Learning

By Weiwei Cheng, Johannes Fürnkranz, Eyke Hüllermeier, Sang-Hyeun Park. 2011

64 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Reinforcement Learning in Healthcare: A Survey (2019, methodology): “wards using qualitative preference or simply based on past actions in the treatment of cancer [89], [52], [98].”
APRIL: Active Preference-learning based Reinforcement Learning (2012, neutral): “Firstly, we considered a continuous action space (the dosage level is a real value in [0 , 1]), whereas the action space contains 4 discrete actions in [6].”
Active Reward Learning (2014, background): “Based on this insight, preference based algorithms allow the expert to rank executions and learn a controller based on these rankings [2, 6].”
Preference-learning based Inverse Reinforcement Learning for Dialog Control (2012, background): “proposed reinforcement learning based on preference-learning, this requires comparisons for each action pair [14].”
A Bayesian Approach for Policy Learning from Trajectory Preference Queries (2012, methodology): “[5] action preferences were introduced into the classification based policy iteration”
Active reward learning with a novel acquisition function (2015, background): “Following this intuition, preference based algorithms allow the expert to rank executions and learn a controller based on these rankings (Akrour et al. 2013; Cheng et al. 2011).”
Preference-based Evolutionary Direct Policy Search (2013, methodology): “[3] tackle the problem of learning policies solely on the basis of pairwise comparisons between trajectories, suggesting that one system behavior is preferred to another one but without committing to precise numerical rewards.”
Preference-based reinforcement learning: evolutionary direct policy search using a preference-based racing algorithm (2014, methodology): “More speciﬁcally, Cheng et al. [7] use a method called label ranking to train a model that ranks actions given a state; their approach generalizes classiﬁcation-based approximate policy iteration [19].”