YAN XU
Director, Numerical Optimization team, SAS Institute Inc
Better Machine Learning Models by Derivative-free Optimization
Optimization is a key component in many machine learning (ML) or artificial intelligence algorithms. Optimization is not only used to fit ML models, but also help to create better models in terms of accuracy and complexity. In this tutorial, we first introduce a number of derivative-free optimization (DFO) methods, which have been successfully used to improve ML models by optimizing their hyperparameters. We then present several real-world ML applications that significantly benefit from those DFO methods.
Optimization is a key component in many machine learning (ML) or artificial intelligence algorithms. Optimization is not only used to fit ML models, but also help to create better models in terms of accuracy and complexity. In this tutorial, we first introduce a number of derivative-free optimization (DFO) methods, which have been successfully used to improve ML models by optimizing their hyperparameters. We then present several real-world ML applications that significantly benefit from those DFO methods.