Scribd
Upload
26 views30 pages

Machine Learning Security with Azure

The document discusses Azure Machine Learning (AML), which combines Microsoft's Azure cloud computing platform with machine learning capabilities. It begins with an introduction to AML by the speaker, Al Noel, and then provides overviews of Azure, machine learning, and how Azure and machine learning are integrated through AML. Examples of predictive analytics applications using AML are also presented.

Uploaded by

springlee
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
26 views30 pages

Machine Learning Security with Azure

The document discusses Azure Machine Learning (AML), which combines Microsoft's Azure cloud computing platform with machine learning capabilities. It begins with an introduction to AML by the speaker, Al Noel, and then provides overviews of Azure, machine learning, and how Azure and machine learning are integrated through AML. Examples of predictive analytics applications using AML are also presented.

Uploaded by

springlee
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd

Azure Machine Learning

(AML)
Al Noel
Principal Consultant
Microsoft Consulting Services
Agenda

1. Intro – Al Noel
2. Azure
3. Machine Learning
4. Azure + Machine Learning =
Azure Machine Learning (AML)
Speaker Today – Al Noel
Some Interesting Stuff About Al (or maybe not  )

 35 years in IT
 Developer NSA
 Data Center Manager at Pentagon
 Trainer
 Visiting Scientist AI Lab MIT
 Production SQL Server DBA
 16+ years at Microsoft, SQL Server
 Collect Pez dispensers
 Daughter’s legal name is Bunny
• Lived in:
Turkey, Thailand, Belgium, California, Alabama,
Arizona, Florida, Georgia, Kentucky, Massachusetts,
Rhode Island, Virginia and Maryland
 Currently live and work in Washington, DC area
 Aspiring data scientist
Microsoft Azure
Azure Cloud Computing Platform
 For building, deploying and managing applications and services through a global
network of Microsoft managed data centers (19) with over a million servers world wide.
$15 billion investment. “All In”
 Provides Platform as a Service (PaaS),Infrastructure as a Service (IaaS) and SaaS.
 Supports many different programming languages, tools and frameworks including both
Microsoft specific and third party software and systems (Oracle, Linux, Java, etc.)
 Uses a specialized operating system to run its “fabric layer” which is a huge cluster.
 Scaling and reliability are controlled by the Microsoft Azure Fabric Controller so the
services and environment do not crash if one of the servers crashes within the Microsoft
datacenter
 New services and features constantly being added – frequent upgrades to existing
services using Agile approaches
 [Link]
 According to BusinessWeek magazine at this point the two leading cloud providers are
Amazon Web Services and Microsoft Azure. All others lag these two.
Free eBooks On Azure
Microsoft Azure Essentials: Fundamentals of Azure

[Link]
Machine Learning
Machine Learning

When was the first machine learning program written?

1959 by Arthur Samuel


Machine Learning
 Originally scientific discipline that explores the construction and study of algorithms that
can learn from data.
 Now – lots of real world application
 Algorithms operate by building a “model” based on inputs and using that to make
predictions or decisions, rather than following only explicitly programmed instructions.
 Three broad categories:

 Supervised learning. The computer is presented with example inputs and their desired outputs, given
by a "teacher", and the goal is to learn a general rule that maps inputs to outputs.
 Unsupervised learning, no labels are given to the learning algorithm, leaving it on its own to find
structure in its input. Unsupervised learning can be a goal in itself (discovering hidden patterns in data)
or a means towards an end. Find useful, non obvious predictive rule
 Reinforcement learning, a computer program interacts with a dynamic environment in which it must
perform a certain goal (such as driving a vehicle), without a teacher explicitly telling it whether it has
come close to its goal or not. Another example is learning to play a game by playing against an
opponent
 Computer systems that become “smarter” with experience
Feed the Model Data and It “Learns”
What You Want to be Able to
Variables Predict

A B C

Training Data 1.2 45 14 N


1.4 43 16 N
1.1 67 3 Y
1.2 55 7 Y
1.2 30 13 N

Test Data 1.3 43 16 Should get N


1.1 65 3 Should get Y
Learning, Machines and People –> Small Chunks

Patrick Winston, Professor MIT


“You can only learn that which you almost
already know.”
Value of ML
 Just a 3% improvement in detecting gift card fraud
resulted in $40 million loss avoidance
 Does not have to be perfect to be valuable
 Microsoft’s Vision for ML
Make machine learning accessible to every enterprise, data scientist, developer,
information worker, consumer, and device anywhere in the world
Really Focused on Predictive Analytics
 Credit scoring first used by mail order business in
1950’s
 3 Fundamental Benefits
 Speed – evaluate millions of customers in seconds
 Accuracy - ore accurate than humans – about 20-30%
 Consistency – a model will always generate same prediction given same set of
data – even a competent human expert will not depending on time of day, mood
or whether hungry or not – lots of evidence of this consistency problem
Some Uses of Predictive Analytics
 Identifying people who don’t pay their taxes
 Calculating probability of having a stroke in next 10 years
 Spotting which credit card transactions are fraudulent
 Selecting suspects in criminal cases
 Deciding which candidate to offer a job to
 Predicting how likely it is that a customer will become bankrupt
 Predicting which customers are likely to defect to rival phone plan when
their contract reaches its end
 Determining what books, music and films you are likely to purchase next
 Forecasting life expectancy
Free Data Science and Machine Learning
Courses
 Coursera
 Stanford University Machine Learning
 [Link]
 Next class runs September 15 – Dec 7
 Johns Hopkins University
 9 course Data Science certificate program
 Runs almost continuously
 [Link]

 [Link]
 Refine your search by clicking on “Data Analysis & Statistics”
Edx Course on Data Science from
Microsoft
Azure Machine Learning
(AML)
Azure Machine Learning
 All in the cloud
 Only need a browser and an Azure account for
development, test and production
 No hardware or software to buy or install
 Came out of Preview and into General Availability
Wednesday February 18th
Azure ML Mitigates Four Challenges
 Expenses
 Huge set-up costs of tools, expertise and compute/storage capacity create
significant barriers to entry
 Siloed Data
 Stove piped and cumbersome data management inhibits access to data

 Disconnected Tools
 Complex and fragmented tools limit participation in exploring data and building
models
 Deployment Complexity
 Many models never provide business value due to problems deploying to
production
Microsoft Azure Machine Learning Studio
Collaborative visual development environment that
enables you to:
 Build
 Test
 Deploy

Predictive Analytics solutions


Inside the ML Studio
Azure Management Portal
Good Prediction Problems
 Adequate amount of data – not necessarily huge
amounts
 Labels
 Labeled examples
 Customer churn: records of loyal customers + records of customer who left
(churn)
 Relevant “features”
 Customer information: age, sex, zip code, previous spending patterns
 Transaction information: amount, previous transactions

 Tolerance for uncertainty


 Does not have to be perfect i.e. fraud detection improvement of 3% = $millions
Variety of Applications of Azure ML
On Line -How to Do Data Science in
Azure
[Link]
Azure Market Place - Various Services to Use

 [Link]
 http
://[Link]/browse/data?category=machine-learnin
g
 For example “Frequently Bought Together” model
[Link]
References
 Book from Apress on Azure ML
“Predictive Analytics with Microsoft Azure Machine Learning - Build and Deploy Actionable Solutions in
Minutes”

 Do not buy the printed version, it is unusable, diagrams are all too small to
read! 
 Only buy the companion ebook in PDF format directly from the Apress site
 [Link]

 Predictive Analytics Introduction in General


 Very good introductory book “Predictive Analytics, Data Mining and Big Data: Myths, Misconceptions and Methods “by
Steven Finlay

 http://
[Link]/Predictive-Analytics-Data-Mining-Misconceptions/dp/1137379278/ref=sr_1_fkmr0_1?s=books&ie=UTF8
&qid=1422294583&sr=1-1-fkmr0&keywords=predictive+analytics+steve+finlay
Free eBook Azure Machine Learning
Microsoft Azure Essentials: Azure Machine Learning

[Link]
Free eBook Azure ML and R
Data Science in the Cloud with Microsoft Azure Machine Learning and R

https://
[Link]/[Link]?WT.mc_id=Social_
ServerCloud_General_TTD&Ocid=OutgoingPromotion_Social_TW_MSCloud_20150407_165635968&linkId=1
3380703

You might also like