A Survey on Offline Reinforcement Learning: Taxonomy, Review, and Open Problems / Rafael Figueiredo Prudencio; Marcos R. O. A. Maximo; Esther Luna Colombini.

By: Contributor(s): Material type: ArticleLanguage: English Publisher: [Place of publication not identified] : Institute of Electrical and Electronics Engineers, 2023Description: 1 online resource (1 article)Content type:
  • text
Media type:
  • computer
Carrier type:
  • online resource
Subject(s): Genre/Form: In: IEEE Transactions on Neural Networks and Learning Systems Vol. 35, no. 8, (2023), pages/article 10237-10257Abstract: With the widespread adoption of deep learning, reinforcement learning (RL) has experienced a dramatic increase in popularity, scaling to previously intractable problems, such as playing complex games from pixel observations, sustaining conversations with humans, and controlling robotic agents. However, there is still a wide range of domains inaccessible to RL due to the high cost and danger of interacting with the environment. Offline RL is a paradigm that learns exclusively from static datasets of previously collected interactions, making it feasible to extract policies from large and diverse training datasets. Effective offline RL algorithms have a much wider range of applications than online RL, being particularly appealing for real-world applications, such as education, healthcare, and robotics. In this work, we contribute with a unifying taxonomy to classify offline RL methods. Furthermore, we provide a comprehensive review of the latest algorithmic breakthroughs in the field using a unified notation as well as a review of existing benchmarks' properties and shortcomings. Additionally, we provide a figure that summarizes the performance of each method and class of methods on different dataset properties, equipping researchers with the tools to decide which type of algorithm is best suited for the problem at hand and identify which classes of algorithms look the most promising. Finally, we provide our perspective on open problems and propose future research directions for this rapidly growing field.
List(s) this item appears in: Open Access Resources
Tags from this library: No tags from this library for this title. Log in to add tags.
Star ratings
    Average rating: 0.0 (0 votes)
No physical items for this record

Publication type: Journal article.

Linked open-access copy: published version.

Open access; no subscription required for the linked open-access version. Unrestricted online access star

With the widespread adoption of deep learning, reinforcement learning (RL) has experienced a dramatic increase in popularity, scaling to previously intractable problems, such as playing complex games from pixel observations, sustaining conversations with humans, and controlling robotic agents. However, there is still a wide range of domains inaccessible to RL due to the high cost and danger of interacting with the environment. Offline RL is a paradigm that learns exclusively from static datasets of previously collected interactions, making it feasible to extract policies from large and diverse training datasets. Effective offline RL algorithms have a much wider range of applications than online RL, being particularly appealing for real-world applications, such as education, healthcare, and robotics. In this work, we contribute with a unifying taxonomy to classify offline RL methods. Furthermore, we provide a comprehensive review of the latest algorithmic breakthroughs in the field using a unified notation as well as a review of existing benchmarks' properties and shortcomings. Additionally, we provide a figure that summarizes the performance of each method and class of methods on different dataset properties, equipping researchers with the tools to decide which type of algorithm is best suited for the problem at hand and identify which classes of algorithms look the most promising. Finally, we provide our perspective on open problems and propose future research directions for this rapidly growing field.

Creative Commons Attribution (licence family recorded by the source; version not supplied in metadata).

https://openalex.org/W4360584316

In English.

Description generated from OpenAlex bibliographic metadata retrieved on 5 October 2026; full text linked from the source.

Course allocation: B. Sc. Artificial Intelligence.

There are no comments on this title.

to post a comment.