Abstract
In machine learning deployment, models are often retrained from scratch when new data becomes available or when requirements change. This paper investigates the practice of recycling previously deployed models as starting points for new model development, demonstrating significant improvements in training efficiency and model performance.
Publication Details
- Journal: Empirical Software Engineering
- Volume: 29, Issue 4
- Pages: 100
- Publisher: Springer
- Year: 2024
Key Contributions
- Empirical analysis of model recycling practices in production ML systems
- Framework for effective model reuse strategies
- Performance and cost-benefit analysis of model recycling
Citation
@article{patel2024post,
title={Post deployment recycling of machine learning models: Don't Throw Away Your Old Models!},
author={Patel, Harsh and Adams, Bram and Hassan, Ahmed E},
journal={Empirical Software Engineering},
volume={29},
number={4},
pages={100},
year={2024},
publisher={Springer}
}