Use of static surrogates in hyperparameter optimization
This work addresses the time-consuming optimization process for neural network developers, but it is incremental as it builds on an existing library.
The paper tackled the problem of hyperparameter and architecture optimization for neural networks by enhancing HyperNOMAD with static surrogates to trigger early stopping and candidate ranking, resulting in reduced resource consumption without compromising solution quality.
Optimizing the hyperparameters and architecture of a neural network is a long yet necessary phase in the development of any new application. This consuming process can benefit from the elaboration of strategies designed to quickly discard low quality configurations and focus on more promising candidates. This work aims at enhancing HyperNOMAD, a library that adapts a direct search derivative-free optimization algorithm to tune both the architecture and the training of a neural network simultaneously, by targeting two keys steps of its execution and exploiting cheap approximations in the form of static surrogates to trigger the early stopping of the evaluation of a configuration and the ranking of pools of candidates. These additions to HyperNOMAD are shown to improve on its resources consumption without harming the quality of the proposed solutions.