Encoding Adaptability of Software Engineering Tools as Algorithm Configuration Problem: A Case Study

Nowadays software is often highly configurable, and the required adaptation is a complex and tedious task when performed manually. Moreover, hand-crafted configurations are often far from optimal. In this paper, we study the software configuration problem in the context of the model comparison tool...

Full description

Saved in:
Bibliographic Details
Published in:2019 34th IEEE/ACM International Conference on Automated Software Engineering Workshop (ASEW) pp. 86 - 89
Main Authors: Basmer, Maike, Kehrer, Timo
Format: Conference Proceeding
Language:English
Published: IEEE 01.11.2019
Subjects:
adaptability
Adaptation models
automated algorithm configuration
Conferences
Context modeling
Encoding
Machine learning algorithms
Optimization
SiDiff
SMAC
Software
Software algorithms
Software engineering
Software engineering tools
Tuning
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Nowadays software is often highly configurable, and the required adaptation is a complex and tedious task when performed manually. Moreover, hand-crafted configurations are often far from optimal. In this paper, we study the software configuration problem in the context of the model comparison tool SiDiff, which needs to be carefully adapted to domain-specific modeling languages used in model-driven engineering. To tackle the configuration challenge, we propose to draw from the field of automated algorithm configuration, a research area which has studied the optimization of parameterizable algorithms for many years and which has gained particular momentum through its applications to hyper-parameter tuning in machine learning. Specifically, we report on ongoing work encoding the adaptability of SiDiff as an algorithm configuration problem which is amenable to a sequential model-based optimization tool known as SMAC. While empirical evaluation results are left for future work, the main goal of this paper is to foster active discussions at the workshop and to collect early feedback on our ongoing research.
DOI:10.1109/ASEW.2019.00035