# Transfer Learning for Improving Model Predictions in Highly Configurable Software

## TL;DR

This paper introduces a transfer learning approach that leverages simulated data to predict the performance of highly configurable software systems, reducing measurement costs while maintaining high accuracy and reliability.

## Contribution

It proposes a cost-aware transfer learning method that uses simulation data to improve performance prediction in complex software configurations.

## Key findings

- Achieves high prediction accuracy on real-world systems.
- Reduces measurement effort through simulation-based training.
- Maintains high model reliability across diverse applications.

## Abstract

Modern software systems are built to be used in dynamic environments using configuration capabilities to adapt to changes and external uncertainties. In a self-adaptation context, we are often interested in reasoning about the performance of the systems under different configurations. Usually, we learn a black-box model based on real measurements to predict the performance of the system given a specific configuration. However, as modern systems become more complex, there are many configuration parameters that may interact and we end up learning an exponentially large configuration space. Naturally, this does not scale when relying on real measurements in the actual changing environment. We propose a different solution: Instead of taking the measurements from the real system, we learn the model using samples from other sources, such as simulators that approximate performance of the real system at low cost. We define a cost model that transform the traditional view of model learning into a multi-objective problem that not only takes into account model accuracy but also measurements effort as well. We evaluate our cost-aware transfer learning solution using real-world configurable software including (i) a robotic system, (ii) 3 different stream processing applications, and (iii) a NoSQL database system. The experimental results demonstrate that our approach can achieve (a) a high prediction accuracy, as well as (b) a high model reliability.