Description: Optimizing AI/ML models require fine tuning hyperparameters and architectures to maximize their predictive capabilities and generalization. The number of items to tune can be highly dimensional and evaluations of the formulated objective function can become costly. Deephyper is a framework that uses Asynchronous Bayesian Optimization to perform two common subcomponents of Automated Machine Learning: hyperparameter optimization and neural architecture search (NAS). This course will be an introduction to using Deephyper for supported AutoML tasks. Using HPC’s to parallelize the search method on multi-node CPU and GPU resources will be demonstrated in addition to the basic features available in Deephyper.