Learning with Analytical Models
This work addresses performance prediction for scientific applications, offering an incremental improvement by integrating analytical and machine learning models to enhance accuracy and adaptability to hardware and workload changes.
The paper tackles performance modeling and prediction for scientific applications by proposing a hybrid approach that combines analytical and machine learning models, resulting in improved prediction accuracy compared to pure machine learning techniques, especially with small training datasets.
To understand and predict the performance of scientific applications, several analytical and machine learning approaches have been proposed, each having its advantages and disadvantages. In this paper, we propose and validate a hybrid approach for performance modeling and prediction, which combines analytical and machine learning models. The proposed hybrid model aims to minimize prediction cost while providing reasonable prediction accuracy. Our validation results show that the hybrid model is able to learn and correct the analytical models to better match the actual performance. Furthermore, the proposed hybrid model improves the prediction accuracy in comparison to pure machine learning techniques while using small training datasets, thus making it suitable for hardware and workload changes.