<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>In this work, a study focusing on proposing generalization metrics for Deep Reinforcement Learning (DRL) algorithms was performed. The experiments were conducted in DeepMind Control (DMC) benchmark suite with parameterized environments. The performance of three DRL algorithms in selected ten tasks from the DMC suite has been analysed with existing generalization gap formalism and the proposed ratio and decibel metrics. The results were presented with the proposed methods: average transfer metric and plot for environment normal distribution. These efforts allowed to highlight major changes in the model’s performance and add more insights about making decisions regarding models’ requirements.</p>
</abstract>

In this work, a study focusing on proposing generalization metrics for Deep Reinforcement Learning (DRL) algorithms was performed. The experiments were conducted in DeepMind Control (DMC) benchmark suite with parameterized environments. The performance of three DRL algorithms in selected ten tasks from the DMC suite has been analysed with existing generalization gap formalism and the proposed ratio and decibel metrics. The results were presented with the proposed methods: average transfer metric and plot for environment normal distribution. These efforts allowed to highlight major changes in the model’s performance and add more insights about making decisions regarding models’ requirements.

Metrics for Assessing Generalization of Deep Reinforcement Learning in Parameterized Environments

Department of Automatic Control and Robotics

Journal of Artificial Intelligence and Soft Computing Research

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

In this work, a study focusing on proposing generalization metrics for Deep Reinforcement Learning (DRL) algorithms was performed. The experiments were...

Metrics for Assessing Generalization of Deep Reinforcement Learning in Parameterized Environments

Published Online: Dec 25, 2023

Page range: 45 - 61

Received: Jun 24, 2023

Accepted: Oct 19, 2023

DOI: https://doi.org/10.2478/jaiscr-2024-0003

Keywords
deep reinforcement learning, optimization, generalization, Sim2Sim transfer, adaptation

© 2024 Maciej Aleksandrowicz et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Metrics for Assessing Generalization of Deep Reinforcement Learning in Parameterized Environments

Published Online: Dec 25, 2023

Page range: 45 - 61

Received: Jun 24, 2023

Accepted: Oct 19, 2023

DOI: https://doi.org/10.2478/jaiscr-2024-0003

Keywordsdeep reinforcement learning, optimization, generalization, Sim2Sim transfer, adaptation

© 2024 Maciej Aleksandrowicz et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Keywords
deep reinforcement learning, optimization, generalization, Sim2Sim transfer, adaptation