Multiworld Testing Machine Learning for Contextual Decision-Making.

Slides:



Advertisements
Similar presentations
EzScoreboard.com A Fully Integrated Administration Service.
Advertisements

Recommender Systems & Collaborative Filtering
XProtect® Expert 2013 Product presentation
All content in this presentation is protected – © 2008 American Power Conversion Corporation Rael Haiboullin System Engineer Change Manager.
Context-aware Query Suggestion by Mining Click-through and Session Data Authors: H. Cao et.al KDD 08 Presented by Shize Su 1.
Scaling Distributed Machine Learning with the BASED ON THE PAPER AND PRESENTATION: SCALING DISTRIBUTED MACHINE LEARNING WITH THE PARAMETER SERVER – GOOGLE,
Probabilistic Methods for Targeted Advertising Max Chickering Microsoft Research.
70-290: MCSE Guide to Managing a Microsoft Windows Server 2003 Environment Chapter 11: Monitoring Server Performance.
Academic Advisor: Prof. Ronen Brafman Team Members: Ran Isenberg Mirit Markovich Noa Aharon Alon Furman.
Yahoo Groups Training November Meeting Yahoo Groups.
It all comes together in our… Quickbix Document Suite for One-click documents at your fingertips!
Barracuda Networks Confidential1 Barracuda Backup Service Integrated Local & Offsite Data Backup.
Maintaining Windows Server 2008 File Services
SQL Server 2008 for Hosting Key Questions to Address How can SQL Server save your costs? How can SQL Server help you increase customer base? How can.
Introduction to Oracle Backup and Recovery
Act! emarketing. 2 Increase the Value of Act! Integrated Marketing Easily access contacts, lookups, groups, and companies right from within Act!
Chapter 5 Roles and features. objectives Performing management tasks using the Server Manager console Understanding the Windows Server 2008 roles Understanding.
Exaquantum R2.50 Enhancements Issue b Copyright © Yokogawa Electric Corporation 25th March, 2009 Exaquantum R2.50 Enhancements.
Introduction to the Enterprise Library. Sounds familiar? Writing a component to encapsulate data access Building a component that allows you to log errors.
Committed to Deliver….  We are Leaders in Hadoop Ecosystem.  We support, maintain, monitor and provide services over Hadoop whether you run apache Hadoop,
Challenges in Information Retrieval and Language Modeling Michael Shepherd Dalhousie University Halifax, NS Canada.
Processing and Analyzing Large log from Search Engine Meng Dou 13/9/2012.
Useful Office Practices Denise Brown, NDLTAP July 23, 2013.
Monitoring Latency Sensitive Enterprise Applications on the Cloud Shankar Narayanan Ashiwan Sivakumar.
Making Your Life Easier Marketing for Tech Transfer October 30, 2013 Shaun Juncal Marketing Manager UCSB Office of Technology & Industry Alliances.
Module 7: Fundamentals of Administering Windows Server 2008.
Use Of ICT In Time Management. Objective This presentation is to help show how people can manage their activities to help have enough time for everything.
Evaluation Methods and Challenges. 2 Deepak Agarwal & Bee-Chung ICML’11 Evaluation Methods Ideal method –Experimental Design: Run side-by-side.
We have developed a GUI-based user interface for Chandra data processing automation, data quality evaluation, and control of the system. This system, known.
Reinforcement Learning for Spoken Dialogue Systems: Comparing Strengths & Weaknesses for Practical Deployment Tim Paek Microsoft Research Dialogue on Dialogues.
Replay Compilation: Improving Debuggability of a Just-in Time Complier Presenter: Jun Tao.
70-290: MCSE Guide to Managing a Microsoft Windows Server 2003 Environment, Enhanced Chapter 11: Monitoring Server Performance.
Contextual Ranking of Keywords Using Click Data Utku Irmak, Vadim von Brzeski, Reiner Kraft Yahoo! Inc ICDE 09’ Datamining session Summarized.
Developer TECH REFRESH 15 Junho 2015 #pttechrefres h Understand your end-users and your app with Application Insights.
#mobileema il. AGEND A 1 Why Mobile Isn’t Just About Design 2 Contextual Marketing for 3 Next Steps Join the with.
IMPLEMENTATION OF OUTLOOK EXPRESS
BOĞAZİÇİ UNIVERSITY DEPARTMENT OF MANAGEMENT INFORMATION SYSTEMS MATLAB AS A DATA MINING ENVIRONMENT.
Virtual Classes Provides an Innovative App for Education that Stimulates Engagement and Sharing Content and Experiences in Office 365 MICROSOFT OFFICE.
Web-Mining …searching for the knowledge on the Internet… Marko Grobelnik Institut Jožef Stefan.
OpenField Consolidates Stadium Data, Provides CRM and Analysis Functions for an Intelligent, End-to-End Solution COMPANY PROFILE : OPENFIELD Founded by.
August 2003 At A Glance The IRC is a platform independent, extensible, and adaptive framework that provides robust, interactive, and distributed control.
Chapter VI What should I know about the sizes and speeds of computers?
+ Logentries Is a Real-Time Log Analytics Service for Aggregating, Analyzing, and Alerting on Log Data from Microsoft Azure Apps and Systems MICROSOFT.
PROPRIETARY  2003 Data Research Analysis & Consultancy Solutions All Rights Reserved. This is achieved by: Improving availability / reducing stock outs.
Built atop SharePoint Online, WorkPoint 365 Offers a Project and Case Management Solution to Boost Business Productivity and Deliver Governance OFFICE.
SharePoint ShortUrl and SharePoint Document Merge SharePoint and Office 365 Add-ins to Increase Productivity and Improve User Adoption! OFFICE 365 APP.
The VERSO Product Returns Portal Incorporates Office 365 Outlook and Excel Add-Ins to Create Seamless Workflow for All Participating Users OFFICE 365 APP.
EGI-InSPIRE RI EGI Webinar EGI-InSPIRE RI Porting your application to the EGI Federated Cloud 17 Feb
Instantly Deliver and Track Training to Learners Anytime, Around the World and on Any Device Within Your Office 365 Environment with LMS365 OFFICE 365.
Enterprise Alert on Microsoft Azure Fully Automates Critical Incident Communication and Transforms It into an Intelligent, Reliable, and Mobile Experience.
Let's build a VMM service template from A to Z in one hour Damien Caro Technical Evangelist Microsoft Central & Eastern Europe
Collection Evaluation and Management: Decision Center Launch & Reporter Update Amanda Schukle Product Manager.
AZURE MACHINE LEARNING Bringing New Value To Old Data SQL Saturday #
Use Cases & User Mocks Customer Call – Users and Objects.
Physical Network (L1-L4)
AI Powered ADS A STEP BY STEP GUIDE TO EXTREME PERSONALIZATION
Discover How Your Business Can Benefit from a Facebook Fanpage
Discover How Your Business Can Benefit from a Facebook Fanpage
9/11/ :55 PM Custom Decision Service Contextual Bandits as Service Microsoft Research NYC © 2014 Microsoft Corporation. All rights reserved.
Azure.
Welcome! Google Chrome or Mozilla Firefox are the Browsers of Choice.
Dev Test on Windows Azure Solution in a Box
MyCloudIT Enables Partners to Drive Their Cloud Profitability Using CSP-Enabled Desktop Hosting Automation with Microsoft Azure and Office 365 MICROSOFT.
Azure Makes Model-Based Test Automation and Risk Optimization Possible from the Cloud MINI-CASE STUDY “Tricentis had a traditional fat client in a 100%
Managing Services with VMM and App Controller
Approaching an ML Problem
Technical Capabilities
Windows Client Assessment Results
Microsoft Virtual Academy
Hospital Management System
Presentation transcript:

Multiworld Testing Machine Learning for Contextual Decision-Making

Contextual Decision-Making User Profile Demographics Location Past Behavior ? User Clicks Story User Reads Story User Returns More Service Makes Money

ML for Contextual Decision-Making Given a particular context, select an action that optimizes the reward observed Great for personalization or situational decisions personalized news content-based interruptions for OS scheduling wellness interventions

Experimentation Multiworld testing: Get the right data first, then experiment offline like crazy Statistically: 1 billion experiments, for the cost of 21 A/B tests Read Recommender Ignored Recommender A/B TestingMultiworld Testing

Results: Personalized >30% lift over editorial

Results: >15% revenue improvement* *Deepak large scale learning workshop

Multiworld Testing Decision Service Goal: Make this easy, fast, automated Modular Supports cycle times from 2 minutes to 2 months Response times fast enough for any application Explore LogLearn Deploy any part of 

Decision Service Exploration

Client Library Makes decisions Located within the application for extremely low latency Supports VW models or generic user-defined functions Performs exploration Several exploration algorithms available ɛ-greedy Softmax Bootstrap Generic Sends data to join service for logging Provides compression for feature vectors

Decision Service Logging

Join Service Joins together all data with the same key that arrives within the specified time window Decision data Observation data Other data to log Two versions available Azure ML Microservice Azure Stream Analytics

Semantics Events Key 1 Events Key 2 duration 10:00 11:00 9: Azure Storage

Decision Service Learning

Azure ML Azure Storage data model Graphical framework to perform offline evaluation or optimization Reader supports reading data from Azure Storage Custom reward functions VW training generates models Adds new data to an existing vw model VW evaluate Evaluates the effect a model would have had based on exploration data Supports vw models or custom user- defined functions

Decision Service Deploy

Command Center Controls high-level settings for applications Register applications Change exploration settings Specify new models to deploy

Summary Multiworld Testing is an efficient approach to finding the optimal policies for contextual decision-making MWT Decision Service is a powerful, modular service designed to make it easy to deploy MWT in many applications