Empirical evaluation methods for multiobjective reinforcement learning algorithms

By Peter Vamplew, Richard Dazeley, Adam Berry, Rustam Issabekov, Evan Dekker. Machine-mediated learning, 2011

349 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/c5ffd0bdbecbeee18232676d2e272ba24b025797

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

In Computational Theory and Mathematics

In Molecular Biology

Citations in the corpus

Multi-objective multi-agent decision making: a utility-based analysis and survey (2019, methodology): “SER is employed in most of the multi-objective planning and reinforcement learning literature [169,177].”
A Survey of Multi-Objective Sequential Decision-Making (2013, methodology): “Shelton (2001) demonstrates this approach with a policy-gradient algorithm, while Vamplew et al. (2011) do the same with the TLO method of Gabor et al. (1998).”
Scalarized multi-objective reinforcement learning: Novel design techniques (2013, methodology): “Recently, [9] proposed empirical evaluation techniques for multi-objective RL, together with a few benchmark instances.”
A Generalized Algorithm for Multi-Objective Reinforcement Learning and Policy Adaptation (2019, background): “Deep Sea Treasure (DST) A classic MORL benchmark [14] in which an agent controls a submarine searching for treasures in a 10 × 11-grid world while trading off time-cost and treasure-value.”
A Deep Reinforcement Learning Perspective on Internet Congestion Control (2019, background): “Multiobjective RL (Vamplew et al., 2011) is a framework for handling the tradeoffs between different reward factors, which in principle could be applied here.”
Policy invariance under reward transformations for multi-objective reinforcement learning (2017, methodology): “[24] as a benchmark problem for single-agent MORL algorithms.”
A practical guide to multi-objective reinforcement learning and planning (2021, background): “, 2011] Resource Gathering 3 Discrete Discrete Known [Vamplew et al., 2011] Linked Rings 2 Discrete Discrete Known [Vamplew et al.”
Linear Fitted-Q Iteration with Multiple Reward Functions (2012, background): “Subsequent developments were made that focussed on the infinite-horizon discounted setting and black-box function approximation techniques (Castelletti et al., 2010; Vamplew et al., 2011).”
A Distributional View on Multi-Objective Policy Optimization (2020, methodology): “We also compute the hypervolume of each found Pareto front; this metric is commonly-used for evaluating MORL algorithms (Vamplew et al., 2011).”
Multi-objective reinforcement learning using sets of pareto dominating policies (2014, methodology): “The Deep Sea Treasure (DST) is proposed by Vamplew et al. (2010) and is a standard MORL benchmark instance.”
Multiobjective Reinforcement Learning: A Comprehensive Overview (2015, methodology): “For example, as indicated in [12] and [27], scalarized Q-learning can be used in a multiple-policy manner by executing repeated runs of the Q-learning algorithm using different parameters.”
Multi-Objective Markov Decision Processes for Data-Driven Decision Support (2016, background): “While our Pareto-based approach makes these three improvements, using Pareto-optimality can still result actions being eliminated unnecessarily; this is illustrated in Figure 1(b).”