New Practical Data Science with Amazon SageMaker

During the course we will learn how to use Amazon SageMaker effectively through real-life ML examples. We will go through the typical data science process from analyzing and visualizing a data set to data preparation and feature design. We will also learn practical aspects of model building, training, fine-tuning and deployment using Amazon SageMaker.

What will you learn during the training?

• Preparing datasets for learning
• Training and evaluation of ML models
• Automatic fine-tuning of ML models
• Preparing ML models for production
• Critical thinking about the models produced by ML

The course is recommended for developers and data scientists.

Module 1: Introduction to machine learning
 Types of ML
 Job Roles in ML
 Steps in the ML pipeline

Module 2: Introduction to data prep and SageMaker
 Training and test dataset defined
 Introduction to SageMaker
 Demonstration: SageMaker console
 Demonstration: Launching a Jupyter notebook

Module 3: Problem formulation and dataset preparation
 Business challenge: Customer churn
 Review customer churn dataset

Module 4: Data analysis and visualization
 Demonstration: Loading and visualizing your dataset
 Exercise 1: Relating features to target variables
 Exercise 2: Relationships between attributes
 Demonstration: Cleaning the data

Module 5: Training and evaluating a model
 Types of algorithms
 XGBoost and SageMaker
 Demonstration: Training the data
 Exercise 3: Finishing the estimator definition
 Exercise 4: Setting hyper parameters
 Exercise 5: Deploying the model
 Demonstration: hyper parameter tuning with SageMaker
 Demonstration: Evaluating model performance

Module 6: Automatically tune a model
 Automatic hyper parameter tuning with SageMaker
 Exercises 6-9: Tuning jobs

Module 7: Deployment / production readiness
 Deploying a model to an endpoint
 A/B deployment for testing
 Auto Scaling
 Demonstration: Configure and test auto scaling
 Demonstration: Check hyper parameter tuning job
 Demonstration: AWS Auto Scaling
 Exercise 10-11: Set up AWS Auto Scaling

Module 8: Relative cost of errors
 Cost of various error types
 Demo: Binary classification cutoff

Module 9: Amazon SageMaker architecture and features
 Accessing Amazon SageMaker notebooks in a VPC
 Amazon SageMaker batch transforms
 Amazon SageMaker Ground Truth  Amazon SageMaker Neo

To complete the course, you need a basic understanding of the Python programming language and the concept of Machine Learning.

After completing the training, we recommend the Amazon SageMaker Studio for Data Scientists course.