Hyperparameter tuning for deep learning in natural language processing
Ahmad Aghaebrahimian, Mark Cieliebak · Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences) · 2019
Deep Neural Networks have advanced rapidly over the past several years. However, it still seems like a black art for many people to make use of them efficiently. The reason for this complexity is that obtaining a consistent and outstanding result from a deep architecture requires optimizing many parameters known as hyperparameters. Hyperparameter tuning is an essential task in deep learning, which can make significant changes in network performance. This paper is the essence of over 3000 GPU hours on optimizing a network for a text classification task on a wide array of hyperparameters. We provide a list of hyperparameters to tune in addition to their tuning impact on the network performance. The hope is that such a listing will provide the interested researchers a mean to prioritize their efforts and to modify their deep architecture for getting the best performance with the least effort.