Machine Learning as a Service

Slides:



Advertisements
Similar presentations
Girish Nathan Misha Bilenko Microsoft Azure Machine Learning How to Work with Large Datasets to Build Predictive Models.
Advertisements

T Sponsors Sameer Chabungbam Principal Program Manager, Microsoft Connector API Apps BizTalk Summit 2015 – London ExCeL London | April 13th & 14th.
Sponsors. Smart Apps with Azure ML CHRIS MCHENRY VP OF TECHNOLOGY, INTEGRO
Meet Jeremy Thake Jeremy recently joined Microsoft as Technical Product Manager for the Visual Studio Developer story for Office 365 development.
The Cloud for Modern WakeUpAndCode.com.
Lecture 8 – Platform as a Service. Introduction We have discussed the SPI model of Cloud Computing – IaaS – PaaS – SaaS.
Building Android Apps with Mobile Services Speaker Name Speaker Title Speaker Company Twitter:
“CLOUD COMPUTING” “CLOUD COMPUTING”. SIMPLE INTRO TO CLOUD COMPUTING (download at beginning of class before viewing) SCROLL DOWN TO 2 ND VIDEO SIMPLE.
Training Workshop Windows Azure Platform. Presentation Outline (hidden slide): Technical Level: 200 Intended Audience: Developers & Technical Decision.
Azure Machine Learning: From design to integration Peter Myers M355.
2015 International TechNet Wiki Summit 2015 Nonki Takahashi Session - Small Basic and TechNet (Small Basic - Part 2)
Matthew Winter and Ned Shawa
Breaking points of traditional approach What if you could handle big data?
A Lap around Azure API Apps. Customer Challenges What are API Apps? Creating and Consuming an API App using VS Adding authentication API Connectors Swagger.
Copyright © New Signature Who we are: Focused on consistently delivering great customer experiences. What we do: We help you transform your business.
1 Divya Jain Oct 10 th, 2014 Big Data Products: Where do I start?
The United States Postal Service processed over 150 billion pieces of mail in 2013—far too much for efficient human sorting. But as recently as 1997,
Naqash Ahmed | Microsoft Student Partner. Naqash Ahmed | Student of Bachelors in Software Engineering Microsoft Student Partner since November.
Real-Time Dashboards on Power BI
Apache Hadoop on Windows Azure Avkash Chauhan
Raju Subba Open Source Project: Apache Spark. Introduction Big Data Analytics Engine and it is open source Spark provides APIs in Scala, Java, Python.
Bhakthi Liyanage SQL Saturday Atlanta 15 July 2017
1/27/2018 5:13 AM © Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN.
IBM Predictive Analytics Virtual Users’ Group Meeting March 30, 2016
4/18/2018 3:49 PM © Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN.
Data Platform and Analytics Foundational Training
Microsoft Virtual Academy
BigDL Deep Learning Library on HDInsight
Welcome! Power BI User Group (PUG)
Build interactive data analysis environments using Apache Spark
About Bill Bill Baer (ˈbɛər)
Microsoft Virtual Academy
9/11/ :51 AM Cloud Roadshow © 2014 Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO.
Microsoft Ignite /11/2018 1:18 AM BRK4017
Developing Hybrid Apps on Microsoft Azure Stack
AI development using Data Science Virtual Machines (DSVM) in Azure
Logo here Module 3 Microsoft Azure Web App. Logo here Module Overview Introduction to App Service Overview of Web Apps Hosting Web Applications in Azure.
Microsoft Ignite /22/2018 3:27 PM BRK2121
Introduction to R Programming with AzureML
Azure ML and Cognitive Services
Data Platform and Analytics Foundational Training
Azure Machine Learning & ML Studio
Python Tools for Visual Studio
Introduction to Azure Machine Learning Studio
The Challenges of moving Document Creation to the Cloud
Server & Tools Business
Data science and machine learning at scale, powered by Jupyter
Microsoft Ignite NZ October 2016 SKYCITY, Auckland.
Orchestration and data movement with Azure Data Factory v2
Microsoft Virtual Academy
Learn how to use and customize the Dynamics AX interactive help system
Modern cloud PaaS for mobile apps, web sites, API's and business logic apps
Microsoft Virtual Academy
Jim Nakashima Program Manager Cloud Tools
TechEd /15/2019 8:08 PM © 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks.
2/17/2019 7:32 AM © 2009 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered.
Office 365 Development.
“CLOUD COMPUTING”.
Delivering great hardware solutions for Windows
Predictive Models with SQL Server Machine Learning Services
Visual Studio Tools for Unity 2.0 Preview
The complete developer's guide to the SkyDrive API
Microsoft Virtual Academy
5/25/2019 © 2014 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks.
Big-Data Analytics with Azure HDInsight
Microsoft Virtual Academy
Microsoft Virtual Academy
Microsoft Virtual Academy
Microsoft Virtual Academy
Presentation transcript:

Machine Learning as a Service 𝑃 𝐴 𝑗 𝐵 = 𝑃 𝐴 𝑗 𝑃(𝐵| 𝐴 𝑗 ) 𝑖=1 𝑁 𝑃 𝐴 𝑖 𝑃(𝐵| 𝐴 𝑖 ) Microsoft Azure ML: Machine Learning as a Service Dmitry Petukhov Ω,𝔘, ℙ #MoscowDataFest

Cluster (on-premises/cloud) Challenge ML as a Service (cloud) Local PC Hybrid Model Cluster (on-premises/cloud) ML Framework scikit learn some library Python / R tools MLlib Mahout Python / R tools Python Runtime Yet Another Runtime Python / R on Spark Dark Magic… Execution Engine Spark MapReduce Spark Resource Management YARN / Apache Mesos YARN / Apache Mesos YARN Local OS Storage Local Disc HDFS / S3 HDFS / S3 HDFS Distributed FS

Agenda Intro <- function() { Hello Data Fest! I need your help } Learn <- function() { Azure ML Overview # +Hello Azure ML Demo Data Science Workflow vs Azure ML Code <- function() { ML Skills Cluster Analysis # Demo 1 Twitter sentiment analysis # Demo 2 Coffee <- function() { Q&A Contacts Agenda

Hello Data Fest! Dmitry Petukhov, Software Architect + Developer, Azure Machine Learning. Introduction Dmitry Petukhov, Software Architect + Developer, Microsoft Certified Professional (C#), Big Data Enthusiast && Coffee Addict Researcher & Developer @ OpenWay Hello Data Fest!

Guiding Principles Data Science is far too complex today Azure Machine Learning. Overview Data Science is far too complex today Math Computer Science Domain Reduce complexity to broaden participation No software to install, only web browser; Possibility to develop without writing line of code; Easy deployment and usage using restfull API; Easy collaboration on Azure ML projects; Visual composition with end2end support for Data Science workflow; Extensible, support for R OSS. Guiding Principles Reference: TechEd 2014 Conference

Azure Machine Learning Azure Machine Learning. Overview Data Azure Machine Learning Consumers Data Model API ML Studio (Web IDE) ML Web Services (REST API Services) Cloud storage Azure Storage Azure Table Hive etc. Excel Manage Workspace: Experiments Datasets Trained models Notebooks Access settings API Azure Marketplace (Applications store) Local storage Upload data from PC… Business Apps Azure ML Gallery (community) Business problem Modeling Deployment Business value Reference: TechEd 2014 Conference

Demo #0: Hello Azure ML! Step 1. Get $200 credit Step 2. Get access to Azure Machine Learning. Overview Step 1. Get $200 credit Sign up for Azure free trial. Step 2. Get access to Azure Portal Demo #0: Hello Azure ML! Step 3. Create Azure ML Workspace Step 4. Go to Azure ML Studio & create ML Experiment Step 5. Publish result

Supervised Learning Flow Part #1 Azure Machine Learning. Azure ML Flow Supervised Learning Flow Part #1 Feature extraction - process attempts to create additional relevant features from the existing raw features in the data, and to increase predictive power to the learning algorithm

Supervised Learning Flow Part #2 Azure Machine Learning. Azure ML Flow Supervised Learning Flow Part #2 Feature selection often increases classification accuracy by eliminating irrelevant, redundant, or highly correlated features. Second, it decreases the number of features which makes model training process more efficient – compute and algorithm performance. This is particularly important for learners that are expensive to train such as support vector machines. Process selects the key subset of original data features in an attempt to reduce the dimensionality of the training problem. PCA & ICA LDA, Fisher root mean square error (RMSE) Source

Azure Machine Learning. Azure ML Flow Source: Azure ML Cheat Sheet Accuracy = (TP+TN)/(P + N): ratio of correctly predicted observations. Using accuracy is only good for symmetric data sets where the class distribution is 50/50. Fraction of correct predicted data Precision = TP/(TP+FP): looks at the ratio of correct positive observations. Recall = TP/(TP+FN): ratio of correctly predicted positive events.

Demo #1: ML Skills Cluster Analysis Azure Machine Learning. Demo k-means clustering aims to partition the n observations into k (≤ n) sets S = {S1, S2, …, Sk} so as to minimize the within-cluster sum of squares (WCSS). where (x1, x2, …, xn) – observations, μi is the mean of points in Si. Demo #1: ML Skills Cluster Analysis Source: Wikipedia

Demo #2: Twitter sentiment analysis Azure Machine Learning. Demo TD-IDF, short for term frequency–inverse document frequency, is a numerical statistic that is intended to reflect how important a word is to a document in a collection or corpus. Demo #2: Twitter sentiment analysis Source: Wikipedia

Restrictions Legislative restrictions Azure platform restrictions Azure Machine Learning. Conclusion Legislative restrictions International & local Azure platform restrictions Max storage volume per account, etc. Azure ML service restrictions Data Max dataset volume: 10 Gb Vector size limitation: 2^64 Throttled policy 20 concurrent request per endpoint Max endpoints count: 10K Black box No debug No Scala, C++, C# No your own right algorithms Restrictions

Killer Features R (quickstart) Python (quickstart) Publishing Azure Machine Learning. Conclusion R (quickstart) Support R models & scripts Python (quickstart) Support Python scripts Jupyter Notebooks in Azure ML Studio Publishing REST API & real-time mode vs batch-mode Azure ML Gallery Share for community Azure Marketplace SaaS store In-the-box integration with… Hive, Azure Storage, Excel, Cortana Analytics Stack Free Start & it’s child age Killer Features

Data Science still too complex today Azure Machine Learning. Conclusion Data Science still too complex today Math Computer Science Domain Reduce complexity to broaden participation No software to install, only web browser; Possibility to develop without writing line of code; Easy deployment and usage using restfull API; Easy collaboration on Azure ML projects; Visual composition with end2end support for Data Science workflow; Extensible, support for R OSS. Nothing has changed

References Start from azure.com/ml Microsoft Machine Learning Blog Azure Machine Learning. Conclusion Start from azure.com/ml Microsoft Machine Learning Blog Azure ML documentation + free online course, videos & books Microsoft Research: Azure for Researchers References

Thank you! © 2015 Dmitry Petukhov All rights reserved. Microsoft Azure and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.

Now or later (send on d.petukhov@outlook.com) Stay connected Azure ML: Machine Learning as a Service Q&A Now or later (send on d.petukhov@outlook.com) Stay connected Facebook: @code.zombi LinkedIn: @dpetukhov Habr: @codezombie All contacts… Read my tech code instinct blog Download presentation from http://0xcode.in/moscow-data-fest or