This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1. 2015 Avanade Inc. All Rights Reserved. >Azure Machine
Learning Avanade Italy Speaker Antimo Musone Net Campus 31 Maggio
2015 1 Technical Architect
2. 2015 Avanade Inc. All Rights Reserved. | 2 About Me Engineer
of II University of Naples In Avanade dal 2006 Technical Architect
Cloud & Mobile Co-Founder Fifth Ingenium
3. 2015 Avanade Inc. All Rights Reserved. Agenda 2 Machine
Learning Overview 1 Avanade Italy 3 Azure Machine Learnig 4
Business Case | 3 Copyright 2014 Avanade Inc. All Rights Reserved.
The Avanade name and logo are registered trademarks in the US and
other countries.
4. 4 Chi Avanade Italy Avanade, nata nel 2000 dalla joint
venture tra Microsoft e Accenture, il principale Technology
Integrator a livello mondiale specializzato nello sviluppo e
nellimplementazione di soluzioni su tecnologia Microsoft per le
grandi aziende. Capacit imprenditoriale Soluzioni per qualsiasi
settore Acceleratore di soluzioni trasversali Specializzati su
piattaforma Microsoft Architetture e asset Risultati realizzati
Piattaforme Enterprise Prodotti allavanguardia Impegno per la
ricerca e sviluppo
5. Avanade Italy: I nostri numeri CAGLIARI Nata nel Settembre
2000 (tra i Paesi fondatori) 6 Uffici 700 dipendenti Fatturato 80
milioni di Euro ROMA
6. What is Machine Learning ?
7. Machine Learning / Predictive Analytics Vision Analytics
Recommenda-tion engines Advertising analysis Weather forecasting
for business planning Social network analysis Legal discovery and
document archiving Pricing analysis Fraud detection Churn analysis
Equipment monitoring Location-based tracking and services
Personalized Insurance Machine learning & predictive analytics
are core capabilities that are needed throughout your business
8. Machine Learning Overview Formal definition: The field of
machine learning is concerned with the question of how to construct
computer programs that automatically improve with experience - Tom
M. Mitchell Another definition: The goal of machine learning is to
program computers to use example data or past experience to solve a
given problem. Introduction to Machine Learning, 2nd Edition, MIT
Press ML often involves two primary techniques: Supervised
Learning: Finding the mapping between inputs and outputs using
correct values to train a model Unsupervised Learning: Finding
patterns in the input data (similar to Density Estimates in
Statistics)
9. Machine Learning Data: A B C D E F G H I J K L M N O P Q R S
T U V W X Y Z Rules, or Algorithms: about, Learning, language
Spelling and sounding builds words Learning about language. Words
build sentences Learning, or Abstraction: Any new understanding
proceeds from previous knowledge.
10. Traditional programming VS Machine Learning Computer Data
Program Output Computer Data Output Program Traditional Programming
Machine Learning
11. No, more like gardening Gardener = You Seeds = Algorithms
Nutrients = Data Plants = Programs
12. Sample Application Web search Computational biology Finance
E-commerce Space exploration Robotics Information extraction Social
networks Debugging [Your favorite area]
13. Types of Learning Supervised (inductive) learning Training
data includes desired outputs Unsupervised learning Training data
does not include desired outputs Semi-supervised learning Training
data includes a few desired outputs
14. Machine Learning Problem Classification or Categorization
Clustering Regression Dimensionality reduction Supervised Learning
Unsupervised Learning DiscreteContinuous
15. Machine Learning Example Predict function F(X) for new
examples X Discrete F(X): Classification Continuous F(X):
Regression F(X) = Probability(X): Probability estimation Given
examples of a function (X, F(X))
16. Machine Learning Example Training: given a training set of
labeled examples {(x1,y1), , (xN,yN)}, estimate the prediction
function f by minimizing the prediction error on the training set
Testing: apply f to a never before seen test example x and output
the predicted value y = f(x) output prediction function Image
feature y = f(x) Apply a prediction function to a feature
representation of the image to get the desired output: F( ) = apple
F( ) = tomato F( ) = dog
17. Supervised Learning 1. Used when you want to predict
unknown answers from answers you already have 2. Data is divided
into two parts: the data you will use to teach the system (data
set), and the data to test the algorithm (test set) 3. After you
select and clean the data, you select data points that show the
right relationships in the data. The answers are labels, the
categories/columns/attributes are features and the values
arevalues. 4. Then you select an algorithm to compute the outcome.
(Often you choose more than one) 5. You run the program on the data
set, and check to see if you got the right answer from the test
set. 6. Once you perform the experiment, you select the best model.
This is the final output the model is then used against more data
to get the answers you need
18. Supervised Learning 1. Car 2. Not Car
19. Unsupervised Learning 1. Used when you want to find unknown
answers mostly groupings - directly from data 2. No simple way to
evaluate accuracy of what you learn 3. Evaluates more vectors,
groups into sets or classifications 4. Start with the data 5. Apply
algorithm 6. Evaluate groups
20. Unsupervised Learning Example 1 example A Example 2 example
B Example 3 example C example A example B example C Example 1
Example 2 Example 3 The clustering strategies have more tendency to
transitively group points even if they are not nearby in feature
space
21. Azure Learning Machine Workflow Data Its all about the
data. Heres where you will acquire, compile, and analyze testing
and training data sets for use in creating Azure Machine Learning
predictive models. Create the model Use various machine learning
algorithms to create new models that are capable of making
predictions based on inferences about the data sets. Evaluate the
model Examine the accuracy of new predictive models based on
ability to predict the correct outcome, when both the input and
output values are known in advance. Accuracy is measured in terms
of confidence factor approaching the whole number one. Refine and
evaluate the model Compare, contrast, and combine alternate
predictive models to find the right combination(s) that can
consistently produce the most accurate results. Deploy the model
Expose the new predictive model as a scalable cloud web service,
one that is easily accessible over the Internet by any web browser
or mobile client. Test and use the model Implement the new
predictive model web service in a test or production application
scenario.
22. Azure Machine Learning algorithms Classification algorithms
These are used to classify data into different categories that can
then be used to predict one or more discrete variables, based on
the other attributes in the dataset. Regression algorithms These
are used to predict one or more continuous variables, such as
profit or loss, based on other attributes in the dataset.
Clustering algorithms These determine natural groupings and
patterns in datasets and are used to predict grouping
classifications for a given variable. Supervised learning
Classification algorithms Regression algorithms Unsupervised
learning Clustering algorithms
23. Steps Recap
24. Deploying a prediction model The deployment of a new
prediction model takes the form of exposing a web service which can
then be invoked via the Representational State Transfer (REST)
protocol. Azure Machine Learning web services can be called via two
different exposed interfaces: Single, request/response-style calls.
Batch style calls, where multiple input records are passed into the
web service in a single call and the corresponding response
contains an output list of predictions for each input record. When
a new machine learning prediction model is exposed on the Web, it
performs the following operations: New input data is passed into
the web service in the form of a JavaScript Object Notation (JSON)
payload. The web service then passes the incoming data as inputs
into the Azure Machine Learning prediction model engine. The Azure
Machine Learning model then generates a new prediction based on the
input data and returns the new prediction results to the caller via
a JSON payload.
25. AzureML Why AzureML? Setting up a Microsoft Azure Account
Setting up a Storage Account Loading Data Setting up an AzureML
Workspace Accessing AzureML Studio AzureML Studio Tour
26. Azure ML Getting Started Option 1 Take advantage of a free
Azure trial offer at
http://azure.microsoft.com/en-us/pricing/free-trial Option 2 Take
advantage of the (free) Azure Machine Learning trial offer at
https://studio.azureml.net/Home How to get started ?
27. Azure ML Getting Started Option 1 Take advantage of a free
Azure trial offer at
http://azure.microsoft.com/en-us/pricing/free-trial Option 2 Take
advantage of the (free) Azure Machine Learning trial offer at
https://studio.azureml.net/Home
28. Azure Machine Learning workspaces Experiments Azure ML
Studio Web Services Datasets Modules
29. Demo Azure Machine Learning Portal Overview
30. Create your first Azure Machine Learning experiment
Defining the problem you want to solve e.g. figure out if you like
certain movie whats another movie you should watch? (movie
recommendation ) Designing a Solution Using AzureML 1. Creating an
experiment 2. Load a Data Set 3. Create the Experiment 4. Add
Transformations and Filters 5. Create the Experiment Path and apply
Algorithms 6. Save and Run the Model 7. Publish the Model 8. Use
the Model Saving and Running Publishing and Accessing
31. Demo Azure Machine Learning Demo Predict ratings for a
given user and movie Recommend movie to a given user Find users
related to a given user Find items related to a given item
32. API examples Difference in Proportions Test Lexicon Based
Sentiment Analysis Forecasting-Exponential Smoothing Forecasting -
ETS+STL Forecasting-AutoRegressive Integrated Moving Average
(ARIMA) Normal Distribution Quantile Calculator Normal Distribution
Probability Calculator Normal Distribution Generator Binomial
Distribution Probability Calculator Binomial Distribution Quantile
Calculator Binomial Distribution Generator Multivariate Linear
Regression Survival Analysis Binary Classifier Cluster Model
datamarket.azure.com
33. The rest is up to you
34. Sell your Azure Machine Learnig Service Publish Azure
Machine Learning Web Service to the Azure Marketplace
http://azure.microsoft.com/en-us/documentation/articles/machine-learning-publish-web-service-to-
azure-marketplace http://datamarket.azure.com Azure
Marketplace
35. Check out the Machine Learning marketplace at
datamarket.azure.com Learn how to on-board to the marketplace off
azure.com at Machine Learning Center Visit the Microsoft booth to
talk to the team Next Steps
36. Avanade Business Case Connected Mine Walk Through ( IOT +
Azure Machine Learning)
37. 2014 Avanade Inc. All Rights Reserved. Thanks! 39
38. 2015 Avanade Inc. All Rights Reserved. Follow us on:
http://www.avanade.com/it-it/home
http://blog.avanade.com/inside-avanade/
https://it-it.facebook.com/avanade.italy
https://twitter.com/avanadeitaly
https://www.linkedin.com/company/avanade-italy?trk=biz-brand-tree-
co-name