Learn about the machine learning pipeline steps that can be conducted in Amazon SageMaker, including model training, evaluation, and deployment. Discover which critical step must be done outside of SageMaker.
Table of Contents
Question
Which step of the ML pipeline cannot be conducted in Amazon SageMaker?
A. Problem formation
B. Evaluating data
C. Training
D. Deployment
Answer
A. Problem formation
Explanation
Problem formation is the one step of the machine learning pipeline that cannot be conducted directly within Amazon SageMaker. While SageMaker provides a powerful end-to-end platform for building, training, and deploying ML models, the initial step of formulating and defining the problem to solve with machine learning must be done outside of SageMaker.
Problem formation involves clearly articulating the business problem, determining if it is well-suited for an ML approach, identifying the type of ML task (e.g. classification, regression, clustering), defining success metrics, and planning the high-level solution architecture. This strategic planning work requires close collaboration between business stakeholders, subject matter experts, data scientists, and ML engineers to align on objectives and design an effective ML solution.
Once the ML problem is thoughtfully formulated, the subsequent pipeline steps can leverage SageMaker’s capabilities:
- Preparing and evaluating data using SageMaker notebooks and built-in algorithms
- Training models using optimized distributed training on managed infrastructure
- Tuning hyperparameters to find the best model configuration
- Deploying the trained model to a scalable hosted environment for inference
So in summary, while SageMaker streamlines the core ML development workflow, the essential first step of problem formation cannot be automated in SageMaker and requires human-driven strategic planning and cross-functional collaboration to lay the foundation for a successful ML initiative.
Machine Learning Foundation EDMLFDv1EN-US assessment question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Machine Learning Foundation EDMLFDv1EN-US assessment and earn Machine Learning Foundation EDMLFDv1EN-US badge.