ML Series - Getting Started with ML using Amazon SageMaker - Part 2

2022年に JTP株式会社 へJoinしましたMukeshです!


Welcome back to Part 2 of our journey into the world of Machine Learning using Amazon SageMaker! In Part 1 of this Series, we dipped our toes into understanding the basics of Machine Learning, the types, and our goal with the Cats vs Dogs dataset. Now, it's time to delve deeper into the features of SageMaker. Excited? Let's jump in!

What is Amazon SageMaker?

Amazon SageMaker is a fully-managed service that enables developers and data scientists to build, train, and deploy machine learning models quickly. SageMaker removes the heavy lifting from each step of the machine learning process, ensuring a smoother workflow for ML practitioners.

Image Credit: Image created using ChatGPT, a language model developed by OpenAI.

Features of Amazon SageMaker

Let's explore the salient features of SageMaker that make it a popular choice for machine learning:

1. Jupyter Notebooks

At the heart of SageMaker's interactive environment are the Jupyter notebook instances. These notebooks are a boon for users, facilitating various tasks such as:

  • Data Analysis and Preprocessing: Simplifying the examination and preparation of data for machine learning.
  • Algorithm Experimentation: Offering a playground for testing and tweaking algorithms.
  • Result Visualization: Making it easy to visualize complex data and model outputs.
  • Collaboration and Sharing: Enabling seamless collaboration and sharing of insights among team members.

Image Credit: Image created using ChatGPT, a language model developed by OpenAI.

The interface is intuitive, and the environment comes pre-configured with most of the popular data science libraries.

2. Built-in Algorithms

A standout feature of SageMaker is its extensive suite of built-in algorithms. These algorithms cover a wide range of ML tasks, including regression, clustering, and deep learning. They're optimized for high performance, which translates to faster training times and often more accurate results compared to standard implementations.

3. Automatic Model Tuning

The performance of machine learning models is heavily influenced by hyperparameters. SageMaker's automatic model tuning, also known as hyperparameter optimization, intelligently searches and identifies the optimal hyperparameters, thereby significantly enhancing model accuracy.

Image Credit: Image created with Canva

4. Training and Hosting

SageMaker's fully-managed infrastructure covers both model training and deployment. This feature allows users to:

  • Scalable Training: Effortlessly train models on large datasets without the complexities of managing infrastructure.
  • Real-time, Low-latency Inference: Once models are trained, deploying them for real-time predictions is straightforward and efficient.

5. Multi-model Endpoints

SageMaker's ability to serve multiple models from a single endpoint is especially beneficial for situations requiring various model versions or distinct models for different objectives. This not only maximizes resource utilization but also optimizes costs.

Image Credit: Image created with Canva

6. Data Labeling

The Amazon SageMaker Ground Truth feature significantly accelerates the creation of high-accuracy training datasets for machine learning. It adeptly integrates human labelers when necessary and employs active learning to reduce data labeling time and costs by up to 70%.

7. Inference

AWS Inferentia devices have been developed by AWS to provide optimal performance for deep learning inference tasks at a minimal price.

The initial version of the AWS Inferentia device enhances the capabilities of Amazon Elastic Compute Cloud (Amazon EC2) Inf1 models, offering up to 2.3 times more throughput and reducing inference costs by as much as 70% when compared to similar Amazon EC2 models.

8. SageMaker Pipelines

SageMaker Pipelines are a purpose-built, user-friendly CI/CD service for machine learning. They ensure the reproducibility, automation, and efficient iteration of ML workflows, a crucial aspect for maintaining consistent quality and efficiency in machine learning projects.

Image Credit: Image created using ChatGPT, a language model developed by OpenAI.


Amazon SageMaker stands as a formidable force in the machine learning ecosystem, offering an extensive array of features designed to simplify and optimize every phase of the ML lifecycle. From data preprocessing in Jupyter notebooks to deploying highly optimized models in production environments, SageMaker provides an array of tools and services catering to both beginners and seasoned professionals.

In the next installment of our series, we'll explore the practical aspects of SageMaker, loading our Cats vs Dogs dataset, and initiating our training process. Until then, keep exploring and happy learning!