Domain 2: ML Model Development¶
Weight: 26% of scored content
This domain covers model selection, training, hyperparameter tuning, evaluation, and versioning.
Topics Covered¶
| Topic | Description |
|---|---|
| SageMaker Training | Training jobs, instance types, distributed training |
| Built-in Algorithms | XGBoost, Linear Learner, etc. |
| Hyperparameter Tuning | Automatic model tuning |
| Model Evaluation | Metrics, SageMaker Clarify |
| Model Versioning | Model Registry |
| Amazon Bedrock | Foundation models |
Key Concepts¶
SageMaker Training Workflow¶
graph LR
A[Training Data in S3] --> B[Training Job]
B --> C[Model Artifacts]
C --> D[Model Registry]
D --> E[Deployment]Choosing the Right Algorithm¶
| Problem Type | Built-in Algorithms |
|---|---|
| Classification | XGBoost, Linear Learner, KNN |
| Regression | XGBoost, Linear Learner |
| Clustering | K-Means |
| Anomaly Detection | Random Cut Forest |
| NLP | BlazingText, Seq2Seq |
| Computer Vision | Image Classification, Object Detection |
| Recommendations | Factorization Machines |
Study Checklist¶
- Understand SageMaker training job configuration
- Know built-in algorithms and their use cases
- Understand hyperparameter tuning strategies
- Know evaluation metrics for different problem types
- Understand Model Registry for versioning
- Know when to use Bedrock vs custom training