Can I read XCS on embedded systems: an analysis of execution profiles and accelerated classifier deletion on EtoBox?

XCS on embedded systems: an analysis of execution profiles and accelerated classifier deletion by Mathis Brede; Tim Hansmeier; Marco Platzner is a scholarly article available to read on EtoBox.

What is XCS on embedded systems: an analysis of execution profiles and accelerated classifier deletion about?

With the more widespread use of Learning Classifier Systems (LCS) in real-world applications, they are increasingly deployed on embedded computing systems with constrained computational power. We profile the most popular classifier system XCS on both an embedded system with an ARM CPU and an off-the-shelf desktop computer. A comparison of both systems yields distinct differences, as classifier deletion is the most runtime consuming function on the embedded system while matching dominates the execution on the desktop computer. To accelerate classifier deletion and make XCS more amenable for deployment on embedded systems, we propose and evaluate two different deletion mechanisms. While the first is an algorithmic modification of the original deletion approach preserving its behavior, the second is a hybrid scheme utilizing both global and local deletion. We find that the latter can achieve a speedup of up to 1.47 for the overall execution of XCS. CCS CONCEPTS • Computing methodologies → Rule learning.

Author

Mathis Brede; Tim Hansmeier; Marco Platzner

Publisher

ACM

Published

2022

Language

EN