Probabilistic policy reuse in a reinforcement learning agent

By Fernando Fernández, Manuela M. Veloso. 2006

356 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

A comprehensive survey on safe reinforcement learning (2015, methodology): “The integration of this advice together with the π-reuse exploration strategy (Fernández and Veloso, 2006; Fernández et al., 2010) results in the PR-SRL algorithm (Garćıa et al.”
Block Contextual MDPs for Continual Learning (2021, background): “…works have looked at multi-task and transfer learning from the perspective of policy transfer (Rusu et al., 2015; Yin & Pan, 2017), policy reuse (Fernández & Veloso, 2006; Barreto et al., 2016), representation transfer (Rusu et al., 2016; Devin et al., 2017; Andreas et al., 2017; Sodhani et…”
Autonomous transfer for reinforcement learning (2008, background): “[4] F. Fernandez and M. Veloso.”
Boosted Curriculum Reinforcement Learning (2022, background): “…transfer of knowledge between tasks can then happen in various forms, e.g., via samples (Lazaric et al., 2008; Lazaric & Restelli, 2011), policies (Fern´andez & Veloso, 2006; Fern´andez et al., 2010; Silver et al., 2018), reward functions (Brys et al., 2015) or demonstrations/teachers (Chernova &…”
Residual Policy Learning (2018, methodology): “This is not unlike policy-reuse methods, which treat the previously learned policy as the expert, and design an (cid:15) -greedy strategy for sampling new policies [32]. and α were determined for our experiments by performing a small grid search to optimize performance for the hook environment.”
A Meta-MDP Approach to Exploration for Lifelong Reinforcement Learning (2019, background): “Another relevant work in exploration was presented by Fernandez and Veloso [4], where the authors propose building a library of policies from prior experience to explore the environment in new problems more efficiently.”
Approximate Value Iteration with Temporally Extended Actions (2015, methodology): “However, more computationally expensive methods for acquiring options (Simsek & Barto, 2004; Mannor et al., 2004) could be justified if the options are reused for planning in many tasks (Fernández & Veloso, 2006).”
Transfer Learning for Reinforcement Learning Domains: A Survey (2009, background): “Probabilistic policy reuse(Fernandez and Veloso, 2006) also considers a distribution of tasks in which only the goal state differs, but is one of the most robust MTL methods interms of appropriate source task selection.”
Transfer Learning via Inter-Task Mappings for Temporal Difference Learning (2007, background): “Policies from different tasks can also be learned and then used to speed up the current task (Fernandez and Veloso, 2006).”
Combining manual feedback with subsequent MDP reward signals for reinforcement learning (2010, background): “Veloso [5] created an agent that uses a policy from a previous task to aid learning in a task that differed only in its reward function.”
Integrating reinforcement learning with human demonstrations of varying ability (2011, methodology): “The third method used is Probabilistic Policy Reuse , based on the π-reuse Exploration Strategy [4, 5].”