An ML specialist has a custom container image used to train an image classifier in Amazon SageMaker and wants to compare hyperparameter optimization (HPO) results from that custom image to results from SageMaker's built-in image classification algorithm. All experiments and HPO jobs must be launched from scripts inside SageMaker Studio notebooks. Which approach will let the specialist do this in the least amount of time?

The correct approach leverages SageMaker Experiments for the custom container and SageMaker automatic model tuning (AMT) for the built-in algorithm. SageMaker Experiments allows you to organize, track, and compare multiple training jobs, which is ideal for manually tuning a custom model by running various configurations and then analyzing their performance. For the built-in algorithm, SageMaker AMT is the most efficient way to find optimal hyperparameters, as it automates the process of launching and managing tuning jobs. This combination provides the most streamlined path to comparing results from both models. Option 1 is incorrect because local mode in SageMaker Studio is not suitable for HPO, as it runs on the local instance and doesn't scale for multiple training jobs or leverage SageMaker's distributed training capabilities. Option 2 is incorrect because SageMaker Autopilot is designed for automated machine learning (AutoML) on tabular data and is not typically used for tuning custom image classification models or comparing directly with HPO results from other methods. Option 4 is incorrect because SageMaker AMT cannot directly tune a custom container model without specific integration, and the question implies the custom container is not set up for direct AMT.