AutoML: the model that picks your model

Automated search really works. But most of the gain comes not from the search algorithm but from who defined the search space.

Sources

• Bergstra, J. & Bengio, Y. 2012 · Random Search for Hyper-Parameter Optimization · JMLR 13
• Feurer, M. et al. 2015 · Efficient and Robust Automated Machine Learning (auto-sklearn) · NeurIPS 2015
• Erickson, N. et al. 2020 · AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data · arXiv:2003.06505
• Yang, L. & Shami, A. 2020 · On Hyperparameter Optimization of Machine Learning Algorithms: Theory and Practice · Neurocomputing 415