Abstract: Make hyperparameters great again
While tuning hyperparameters of machine learning algorithms is computationally expensive, it also proves vital for improving their predictive performance. Methods for tuning range from manual search to more complex procedures like Bayesian optimization. This talk will demonstrate the latest methods for finding good hyperparameter-sets within a set period of time for common algorithms like xgboost.
Daniel Kühn is a catastrophe analyst at Guy Carpenter, one of the world’s largest reinsurance brokers. In his work he uses state of the art probabilistic models to estimate the economic damage caused by large scale natural disasters. He also is a part time PhD student at the department for computational statics at LMU, where he focuses his research on hyperparameter optimization and automatic machine learning.