Model Performance Prediction for Hyperparameter Optimization of Deep Learning Models Using High Performance Computing and Quantum Annealing
This addresses the problem of reducing computational costs in hyperparameter optimization for researchers and practitioners in fields like high-energy physics, computer vision, and natural language processing, though it is incremental as it builds on existing methods like Hyperband.
The paper tackles the compute-intensive nature of hyperparameter optimization for deep learning models by integrating model performance prediction with early stopping, resulting in Swift-Hyperband, which finds comparable or better hyperparameters while using less computational resources across various domains.
Hyperparameter Optimization (HPO) of Deep Learning-based models tends to be a compute resource intensive process as it usually requires to train the target model with many different hyperparameter configurations. We show that integrating model performance prediction with early stopping methods holds great potential to speed up the HPO process of deep learning models. Moreover, we propose a novel algorithm called Swift-Hyperband that can use either classical or quantum support vector regression for performance prediction and benefit from distributed High Performance Computing environments. This algorithm is tested not only for the Machine-Learned Particle Flow model used in High Energy Physics, but also for a wider range of target models from domains such as computer vision and natural language processing. Swift-Hyperband is shown to find comparable (or better) hyperparameters as well as using less computational resources in all test cases.