Running and managing multiple experiments with SageMaker Experiments

Managing a single machine learning (ML) experiment is easy. When we are dealing with a single ML experiment, it is easy to locate and audit the input and output artifacts, configuration parameters, hyperparameter values, and all the other relevant metadata and details related to this single ML experiment. Things get a bit trickier when we have to deal with multiple ML experiments as well as when retrieving information on experiments and training jobs performed in the past.

In this recipe, we will run and track multiple experiments using SageMaker Experiments. Each experiment trial corresponds to a specific combination of hyperparameters that we will use for the training job. We will use the XGBoost built-in algorithm to help us train and build a classifier using the synthetic dataset we generated in the Synthetic data generation for classification problems recipe. While setting up the experiment, we will make...