Module 2

AW S A c a d e m y M a c h i n e L e a r n i n g F o u n d a t i o n s
Module 2: Introduction to Machine

Learning
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Module overview
Sections Demonstration
Introducing Amazon SageMaker
1. What is machine learning?
2. Business problems solved with
machine learning
3. Machine learning process
4. Machine learning tools overview
5. Machine learning challenges
Knowledge check
© 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved. 2
Module objectives
At the end of this module, you should be able to:

• Recognize how machine learning and deep learning are part of artificial intelligence
• Describe artificial intelligence and machine learning terminology
• Identify how machine learning can be used to solve a business problem
• Describe the machine learning process
• List the tools available to data scientists
• Identify when to use machine learning instead of traditional software development
methods
Module 2: Introduction to Machine Learning
Section 1: What is machine learning ?
Artificial intelligence, machine learning, and
deep learning
Artificial intelligence
Machine learning
Deep learning
Artificial intelligence
Machine learning
Machine learning is the scientific study of algorithms and statistical models to

perform a task using inference instead of instructions.
Data Model Prediction
Machine learning flow
Deep learning
ML and technology advancements
Traditional Cloud computing and Modern machine

computing Moore’s law learning
• Artificial intelligence
Section 1 key • Machines performing human tasks
takeaways • Machine learning
• Training models to make predictions
• Deep learning
• Neural networks
• Technology and economic
advancements have made machine
learning more accessible to
individuals and organizations
10 © 2020, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Section 2: Business problems solved with machine

learning
Common business use cases
Recommended items
Spam versus Recommendations Fraud

regular email
Types of machine learning
Supervised Machine Reinforcement

learning learning learning
Unsupervised
learning
Supervised learning
Learn by identifying patterns in data that is already labeled.
Binary [0,1] • Fraud detection
• Image recognition
Classification
• Customer retention
Multi [0,1,2] • Medical diagnostics
Supervised • Personalized advertising
learning
• Product sales prediction
Regression • Weather forecasting
• Market forecasting
• Population growth prediction
Computer vision
Unsupervised learning
The machine must uncover and create the labels itself.
• Product recommendations
• Customer segmentation
Clustering
• Targeted marketing
• Medical diagnostics
Unsupervised
learning
• Visualization
Dimensionality • Natural language processing
reduction • Data structure discovery
• Gene sequencing
Natural language processing
gögn eru lykilatriði los datos son clave

Hi!
I’d like two

chickens,
please. Once upon a
time
Right, two
chickens, let me
get that started for
you.
Poor story. Little character development. Jumps between scenes like you
might get caught stealing. Unexplained bad guys appear with the thinnest of
back story. Back to unlimited resources and lets not talk about the
mechanics of building something so huge in such an inhospitable place...
Reinforcement learning
Learning through trial and error.
• Game AI
Reinforcement • Self-driving cars
learning
• Robotics +1+
1
• Customer service routing
AWS DeepRacer
Best when the desired outcome is known but the
exact path to achieving it is not known.
Self-driving vehicles
When to use machine learning?
Classical programming approach

Use machine learning when you have:
 Large datasets, large number of variables
Business logic  Lack of clear procedures to obtain the solution
Task
Procedures  Existing machine learning expertise
 Infrastructure already in place to support ML
 Management support for ML
Machine learning approach
Data Model Task
• Machine learning applications affect
Section 2 key everyday life
takeaways • Machine learning can be grouped
into –
• Supervised learning
• Unsupervised learning
• Reinforcement learning
• Most problems are supervised
learning
Section 3: Machine learning process
ML pipeline: Business problem
Business problem
Problem
formulation
ML pipeline: Data preparation
Data handling and cleaning
Business problem
data
Problem data data

formulation
data
Collect and
Evaluate data
Name Country Sex dob
label data
Richard Roe UK Male 18/2/1972
Paulo Santos Male 11/2/1969
Mrs. Mary Major Denver F 37
Desai, Arnav USA M 2/22/1962
ML pipeline: Iterative model training
Business problem
Problem
Tune model
formulation
Meets
Collect and Feature Select and train
Evaluate data Evaluate model business
label data engineering model
goal?
No
Feature augmentation
Data augmentation
ML pipeline: Feature engineering
Name Country Sex dob
Richard Roe UK Male 18/2/1972
Paulo Santos Male 11/2/1969
?
Mrs. Mary Major Denver F 37
Desai, Arnav USA M 2/22/1962
Name USA UK sex age bm dow target

Richard Roe 0 1 0 49 2 5 140,000
Paulo Santos 1 0 0 51 11 7 78,000
Mary Major 1 0 1 37 NAN 0 167,000
Arnav Desai 1 0 0 58 2 4 100,000
ML pipeline: Model training

10–20%
Richard Roe 0 1 0 49 2 5 140,000 Test data
… … … … … … … …
80%
Algorithm
Trained model
XGBoost
{hyperparameters}
ML pipeline: Evaluating and tuning the
model
10-20%
Richard Roe 0 1 0 49 2 5 140,000 Test data
… … … … … … … …
Change 80% predict

features
Algorithm
Trained model Hosted model
XGBoost
{hyperparameters}
Change hyperparameters
Metrics
Overfitting and underfitting
Y Y Y
X X X
Overfitting Underfitting Balanced
ML pipeline: Deployment
New data, retraining
Deploy model
Business problem
Problem Yes
Tune model
formulation
Meets
Collect and Feature Select and train
Evaluate data Evaluate model business
label data engineering model
goal?
No
Feature augmentation
Data augmentation
• Machine learning pipeline guides
Section 3 key you through the process of
takeaways evaluating and training a model
• Iterative process of –
• Data processing
• Training
• Evaluation
Section 4: Machine learning tools overview
Python tools and libraries
• Jupyter Notebook • pandas

• JupyterLab • Matplotlib
• Seaborn
• NumPy
• scikit-learn
Machine learning frameworks and
infrastructure
Machine learning frameworks provide Amazon instances that are designed for
tools and code libraries: machine learning applications:
• Customized scripting • AWS IoT Greengrass provides an
• Integration with AWS services infrastructure for building machine learning
for IoT devices
• Community of developers
• Amazon Elastic Inference reduces costs for
running machine learning applications
PyTorch Caffe2 Torch
TensorFlow Gluon Chainer EC2 P3 EC2 C5 AWS IoT Amazon

instances and C5n Greengrass Elastic
instances Inference
Keras CNTK Apache MXNet
Amazon SageMaker
Ground Truth Notebook Training Inference AWS Marketplace

Set up and manage Provide AWS and Train and tune Create models from Find, buy, and deploy
labeling jobs for SageMaker SDKs models at any scale. training jobs, or import ready-to-use model
highly accurate and sample Use high- external models for packages, algorithms,
training datasets by notebooks to create performance AWS hosting so you can run and data products in
using active learning training jobs and algorithms, or bring inferences on new data. AWS Marketplace.
and human labeling. deploy models. your own.
Demonstration:
Introducing
Amazon
SageMaker
Machine learning managed services
These managed services don’t require ML experience.

Computer Chatbots
vision
Amazon Rekognition Amazon Textract Amazon Lex
Speech Forecasting
Amazon Polly Amazon Transcribe Amazon Forecast
Language Recommendations
Amazon Comprehend Amazon Translate Amazon Personalize

• Python is the most popular ML
Section 4 key language
takeaways • Jupyter Notebooks
• Many open-source tools
• Frameworks and services for all
requirements
• Low-level frameworks
• Amazon SageMaker
• Managed ML services
Section 5: Machine learning challenges
Machine learning challenges
• Poor quality
• Lack of data science expertise
• Non-representative
• Cost of staffing with data scientists
• Insufficient
Data Users • Lack of management support
• Overfitting and underfitting
• Complexity in formulating questions • Data privacy issues

• Explaining models to the business • Tool selection can be complicated
• Cost of building systems • Integration with other systems
Business Technology
Using existing models and services
• Amazon ML managed services

• No ML experience needed
Amazon ML
managed services
You Only Look • Use existing trained and tuned models

Once • Enhance with domain-specific instances
(YOLO)
• Over 250 ML model packages and algorithms

• Over 14 industry segments
AWS Marketplace
• Machine learning challenges
Section 5 key • Data
takeaways • People
• Business
• Technology
• Managed services simplify machine
learning
Module wrap-up
• Machine learning is a subset of artificial intelligence
• Machine learning applies learning algorithms to
Module takeaways develop models from large datasets
• The machine learning pipeline describes the
different stages for developing a machine learning
application
• The Amazon Machine Learning stack has three key
layers
• Managed services, machine learning services,
machine learning frameworks
• Machine learning development is different from
traditional development
• Training algorithm is applied to data to create a
model for making predictions
Module summary
In summary, in this module, you learned how to:

• Recognize how machine learning and deep learning are part of artificial intelligence
• Describe artificial intelligence and machine learning terminology
• Identify how machine learning can be used to solve a business problem
• Describe the machine learning process
• List the tools available to data scientists
• Identify when to use machine learning instead of traditional software development
methods
Complete the knowledge check
Additional resources
• What is Machine Learning?

• Machine Learning on AWS
Thank you
© 2021 Amazon Web Services, Inc. or its affiliates. All rights reserved. This work may not be reproduced or redistributed, in whole or in part, without prior written permission from Amazon
Web Services, Inc. Commercial copying, lending, or selling is prohibited. Corrections, feedback, or other questions? Contact us at https://support.aws.amazon.com/#/contacts/aws-training.
All trademarks are the property of their owners.

Module 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Module 2

Uploaded by

Copyright:

Available Formats

AW S A c a d e m y M a c h i n e L e a r n i n g F o u n d a t i o n s

Module 2: Introduction to Machine

At the end of this module, you should be able to:

Section 1: What is machine learning ?

Machine learning is the scientific study of algorithms and statistical models to

Data Model Prediction

Machine learning flow

Traditional Cloud computing and Modern machine

Section 2: Business problems solved with machine

Spam versus Recommendations Fraud

Supervised Machine Reinforcement

The machine must uncover and create the labels itself.

gögn eru lykilatriði los datos son clave

I’d like two

Learning through trial and error.

Classical programming approach

Data Model Task

Section 3: Machine learning process

Problem data data

Name USA UK sex age bm dow target

Name USA UK sex age bm dow target

Change 80% predict

Section 4: Machine learning tools overview

• Jupyter Notebook • pandas

PyTorch Caffe2 Torch

TensorFlow Gluon Chainer EC2 P3 EC2 C5 AWS IoT Amazon

Ground Truth Notebook Training Inference AWS Marketplace

These managed services don’t require ML experience.

Amazon Rekognition Amazon Textract Amazon Lex

Amazon Polly Amazon Transcribe Amazon Forecast

Amazon Comprehend Amazon Translate Amazon Personalize

Section 5: Machine learning challenges

• Complexity in formulating questions • Data privacy issues

• Amazon ML managed services

You Only Look • Use existing trained and tuned models

• Over 250 ML model packages and algorithms

In summary, in this module, you learned how to:

• What is Machine Learning?

You might also like