Optimizing ML
To uncover how we can optimize our ML costs, we must first understand which tasks constitute an ML workflow. We’ll look at the various steps involved in a typical ML process. Then, we’ll apply optimization methods to those specific steps using the various capabilities in AWS. We’ll focus on how you can optimize your model-training costs and model-deployment costs with Amazon SageMaker.
Understanding an ML workflow
An ML workflow typically requires data exploration and then feature engineering (FE) to transfer data to a format that can be used by an ML algorithm. The algorithm reads the data to find patterns and learns in a sense to generalize patterns so that it can predict outcomes on new, or unknown, data. This is often referred to as model training—you’re applying some mathematical algorithm that may be known and used popularly or something you created yourself to data that is proprietary to you or your organization. The application...