What is it about?

We propose an approach to test the plasticity of RL-based systems, i.e. their ability to adapt to scenarios that may deviate from the training ones. The output of our approach is the frontier of adaptation together with the frontier of regression of the RL system, describing the trade-off between adapting to new scenarios without forgetting the training ones.

Featured Image

Why is it important?

No study has so for explored the adaptation capabilities of a RL-based system. Studying the adaptation is essential to understand the limitations of the system and to decide whether online learning can be safely enabled or not.

Read the Original

This page is a summary of: Testing the Plasticity of Reinforcement Learning Based Systems, ACM Transactions on Software Engineering and Methodology, March 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3511701.
You can read the full text:




The following have contributed to this page