This document provides an overview of a presentation on Amazon SageMaker. The presentation agenda includes a quick overview of SageMaker, followed by labs demonstrating how to load data from S3, train and deploy models with built-in algorithms, perform hyperparameter tuning, and make predictions. The presentation also briefly previews deep learning with SageMaker. The goal is to predict customer enrollment in term deposits using the XGBoost algorithm on a banking dataset.

1. S U M M I T
TORONTO

2. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Build, train, and deploy ML models
withAmazonSageMaker
Jonathan Dion
Senior Technical Evangelist
Amazon Web Services
jondion@amazon.com
@jotdion
linkedin.com/in/jotdion
A I M 3 0 2

3. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Agenda
• Welcome & housekeeping
• Slides: quick overview of Amazon SageMaker
• Labs
• What we’ll cover today:
• Loading data from Amazon Simple Storage Service (Amazon S3)
• Training and deploying with built-in algorithms
• Finding optimal hyperparameters with Automatic ModelTuning
• Running HTTPS predictions and batch predictions
• Beyond built-in algorithms: a peek at Deep Learning

4. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

5. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Housekeeping
• Please be a good neighbor 
• Turn off network backups and any network-hogging apps
• Switch your phones to silent mode
• Help the people around you if you can

6. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

7. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
AmazonSageMaker
Collect and prepare
training data
Choose and optimize
your ML algorithm
Set up and manage
environments for
training
Train and tune model
(trial and error)
Deploy model
in production
Scale and manage the
production
environment
Easily build, train, and deploy machine learning models

8. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
AmazonSageMaker
Notebook
instances
K-Means Clustering
Principal Component Analysis
Neural Topic Modelling
Factorization Machines
Linear Learner
XGBoost
Latent Dirichlet Allocation
Image Classification
Seq2Seq,
And more!
ALGORITHMS
Apache MXNet, Chainer
TensorFlow, PyTorch
Caffe2, CNTK,
Torch
FRAMEWORKS Set up and manage
environments for training
Train and tune
model (trial and
error)
Deploy model
in production
Scale and manage the
production environment
Built-in, high-
performance
algorithms
Build

9. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
AmazonSageMaker
Notebook
instances
Built-in, high-
performance
algorithms
One-click
training
Automatic
Model Tuning
Build Train
Deploy model
in production
Scale and manage the
production
environment

10. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
AmazonSageMaker
Fully managed hosting
with auto-scaling
One-click
deployment
Notebook
instances
Built-in, high-
performance
algorithms
One-click
training
Automatic
Model Tuning
Build Train Deploy

11. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Model Training (on EC2)
Model Hosting (on EC2)
Trainingdata
Modelartifacts
Training code Helper code
Helper codeInference code
GroundTruth
Client application
Inference code
Training code
Inference requestInference response
Inference Endpoint

12. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Training Code
Factorization Machines
Linear Learner
Principal Component Analysis
K-Means
XGBoost
And more
Built-in Algorithms BringYour Own ContainerBringYour Own Script
Model options

13. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
AmazonSageMakerSDK
• Python SDK orchestrating allAmazon SageMaker activity
• Algorithm selection, training, deploying, hyperparameter optimization, and so on
• There’s also a Spark SDK (Python and Scala) which we won’t cover today
• High-level objects for:
• Some built-in algos: Kmeans, PCA, and the like
• Deep learning libraries:TensorFlow, MXNet, PyTorch, Chainer
• Sagemaker.estimator.estimator for everything else
https://github.com/aws/sagemaker-python-sdk
https://sagemaker.readthedocs.io/en/latest/

14. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Built-in algorithms
pink:supervised,blue:unsupervised
Linear learner: regression, classification Image classification: deep learning (ResNet)
Factorization machines: regression, classification,
recommendation
Object detection: deep learning
(VGG or ResNet)
K-Nearest neighbors: non-parametric regression and
classification
Neural topic model: topic modeling
XGBoost: regression, classification, ranking
https://github.com/dmlc/xgboost
Latent Dirichlet allocation: topic modeling (mostly)
K-Means: clustering Blazing text: GPU-based Word2Vec,
and text classification
Principal component analysis: dimensionality reduction Sequence to sequence: machine translation, speech-to-text
and more
Random cut forest: anomaly detection DeepAR: time-series forecasting (RNN)
Object2Vec general-purpose embeddings IP Insights: usage patterns for IP addresses

15. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
XGBoost
• Open Source project
• Popular tree-based algorithm
for regression, classification and
ranking
• Handles missing values
and sparse data
• Supports distributed training
• Can work with data sets larger
than RAM
https://github.com/dmlc/xgboost
https://xgboost.readthedocs.io/en/latest/
https://arxiv.org/abs/1603.02754

16. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Confusion matrix
True
Negative
False
Positive
False
Negative
True
Positive
Actual
Predict
0
1
0 1
3407 231
257 224
Actual
Predict
0
1
0 1

17. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Loading training datafromAmazonS3
• Two modes: File Mode and Pipe Mode
• input_mode parameter in sagemaker.estimator.estimator
• File Mode copies the data set to training instances
• You need to provision enough storage
• S3DataSource object
• S3DataDistributionType : FullyReplicated | ShardedByS3Key
• Different data formats are supported: CSV, protobuf, JSON, libsvm (check algo docs!)
• Pipe Mode streams the data set to training instances
• This allows you to process infinitely-large data sets
• Training starts faster
• This mode is supported by some built-in algos as well asTensorFlow
• Your data set must be in recordIO-encoded protobuf format.

18. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

19. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
ProblemStatement
Direct marketing is a common tactic to acquire customers. Because
resources and a customer's attention is limited, the goal is to only target the
subset of prospects who are likely to engage with a specific offer.
Predicting those potential customers based on readily available information
like demographics, past interactions, and environmental factors is a
common machine learning problem.
We will train a model using XGBoost on a Bank Marketing Dataset provided
by UCI’s ML Repository to predict if a customer will enroll for a term deposit
at a bank, after one or more phone calls.

20. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

21. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Labs
1. Preparing the data
2. Training our first model with XGBoost
3. Deploying our model
4. Predicting with our model using HTTPS
5. Running batch predictions with our model
6. Manually tuning our model
7. Finding optimal hyperparameters with Automatic ModelTuning
8. Deploying our new model

22. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

23. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.S U M M I T
Resources
https://ml.aws
https://aws.amazon.com/sagemaker
https://github.com/awslabs/amazon-sagemaker-examples
https://github.com/aws/sagemaker-python-sdk
https://github.com/awslabs/amazon-sagemaker-workshop

24. Thank you!
S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Jonathan Dion
SeniorTechnical Evangelist
AmazonWeb Services
@jotdion
linkedin.com/in/jotdion

25. S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Please complete the session survey
in the mobile app.
Complete three surveys, and you’ll
receive a gift at the Help Desk.