http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/javascripts/application-5a24945b.js.gz ---------------------------------------------------------------------- diff --git a/javascripts/application-5a24945b.js.gz b/javascripts/application-5a24945b.js.gz index 5066a9d..1469cc9 100644 Binary files a/javascripts/application-5a24945b.js.gz and b/javascripts/application-5a24945b.js.gz differ
http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/javascripts/tryit-b5b64753.js.gz ---------------------------------------------------------------------- diff --git a/javascripts/tryit-b5b64753.js.gz b/javascripts/tryit-b5b64753.js.gz index f2670f7..9e5bc18 100644 Binary files a/javascripts/tryit-b5b64753.js.gz and b/javascripts/tryit-b5b64753.js.gz differ http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/machinelearning/dimensionalityreduction/index.html ---------------------------------------------------------------------- diff --git a/machinelearning/dimensionalityreduction/index.html b/machinelearning/dimensionalityreduction/index.html index 70d204d..5880e38 100644 --- a/machinelearning/dimensionalityreduction/index.html +++ b/machinelearning/dimensionalityreduction/index.html @@ -1,4 +1,4 @@ -<!DOCTYPE html><html><head><title>Dimensionality Reduction With PredictionIO</title><meta charset="utf-8"/><meta content="IE=edge,chrome=1" http-equiv="X-UA-Compatible"/><meta name="viewport" content="width=device-width, initial-scale=1.0"/><meta class="swiftype" name="title" data-type="string" content="Dimensionality Reduction With PredictionIO"/><link rel="canonical" href="https://docs.prediction.io/machinelearning/dimensionalityreduction/"/><link href="/images/favicon/normal-b330020a.png" rel="shortcut icon"/><link href="/images/favicon/apple-c0febcf2.png" rel="apple-touch-icon"/><link href="//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet"/><link href="//maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet"/><link href="/stylesheets/application-a2a2f408.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min. js"></script><script src="//cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}catch(e){}</script></head><body><div id="global"><header><div class="container" id="header-wrapper"><div class="row"><div class="col-sm-12"><div id="logo-wrapper"><span id="drawer-toggle"></span><a href="#"></a><a href="http://predictionio.incubator.apache.org/";><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="//templates.prediction.io/">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-predictionio/">OPEN SOURCE</a></div></div><img class="mobile-search-bar-toggler hidden-md hidden-lg" src="/images/icons/search-glass-704bd4ff.png"/></div></div></div></header><div id="search-bar-row-wrapper"><div class="container-fluid" id="search-bar-row"><div class="row"> <div class="col-md-9 col-sm-11 col-xs-11"><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>Dimensionality Reduction With PredictionIO</h4></div><h4 class="hidden-sm hidden-xs">PredictionIO Docs</h4></div><div class="col-md-3 col-sm-1 col-xs-1 hidden-md hidden-lg"><img id="left-menu-indicator" src="/images/icons/down-arrow-dfe9f7fe.png"/></div><div class="col-md-3 col-sm-12 col-xs-12 swiftype-wrapper"><div class="swiftype"><form class="search-form"><img class="search-box-toggler hidden-xs hidden-sm" src="/images/icons/search-glass-704bd4ff.png"/><div class="search-box"><img src="/images/icons/search-glass-704bd4ff.png"/><input type="text" id="st-search-input" class="st-search-input" placeholder="Search Doc..."/></div><img class="swiftype-row-hider hidden-md hidden-lg" src="/images/icons/drawer-toggle-active-fcbef12a.png"/></form></div></div><div class="mobile-left-menu-toggler hidden-md hidden-lg"></div></div></div></div><div id="page" cla ss="container-fluid"><div class="row"><div id="left-menu-wrapper" class="col-md-3"><nav id="nav-main"><ul><li class="level-1"><a class="expandible" href="/"><span>Apache PredictionIO (incubating) Documentation</span></a><ul><li class="level-2"><a class="final" href="/"><span>Welcome to Apache PredictionIO (incubating)</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Started</span></a><ul><li class="level-2"><a class="final" href="/start/"><span>A Quick Intro</span></a></li><li class="level-2"><a class="final" href="/install/"><span>Installing Apache PredictionIO (incubating)</span></a></li><li class="level-2"><a class="final" href="/start/download/"><span>Downloading an Engine Template</span></a></li><li class="level-2"><a class="final" href="/start/deploy/"><span>Deploying Your First Engine</span></a></li><li class="level-2"><a class="final" href="/start/customize/"><span>Customizing the Engine</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Integrating with Your App</span></a><ul><li class="level-2"><a class="final" href="/appintegration/"><span>App Integration Overview</span></a></li><li class="level-2"><a class="expandible" href="/sdk/"><span>List of SDKs</span></a><ul><li class="level-3"><a class="final" href="/sdk/java/"><span>Java & Android SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/php/"><span>PHP SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/python/"><span>Python SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/ruby/"><span>Ruby SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/community/"><span>Community Powered SDKs</span></a></li></ul></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Deploying an Engine</span></a><ul><li class="level-2"><a class="final" href="/deploy/"><span>Deploying as a Web Service</span></a></li><li class="level-2"><a class="final" href="/cli/#engine-com mands"><span>Engine Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</span></a></li><li class="level-2"><a class="final" href="/deploy/engineparams/"><span>Setting Engine Parameters</span></a></li><li class="level-2"><a class="final" href="/deploy/enginevariants/"><span>Deploying Multiple Engine Variants</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Customizing an Engine</span></a><ul><li class="level-2"><a class="final" href="/customize/"><span>Learning DASE</span></a></li><li class="level-2"><a class="final" href="/customize/dase/"><span>Implement DASE</span></a></li><li class="level-2"><a class="final" href="/customize/troubleshooting/"><span>Troubleshooting Engine Development</span></a></li><li class="level-2"><a class="final" href="/api/current/#package"><span>Engine Scala APIs</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Collecti ng and Analyzing Data</span></a><ul><li class="level-2"><a class="final" href="/datacollection/"><span>Event Server Overview</span></a></li><li class="level-2"><a class="final" href="/cli/#event-server-commands"><span>Event Server Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventapi/"><span>Collecting Data with REST/SDKs</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventmodel/"><span>Events Modeling</span></a></li><li class="level-2"><a class="final" href="/datacollection/webhooks/"><span>Unifying Multichannel Data with Webhooks</span></a></li><li class="level-2"><a class="final" href="/datacollection/channel/"><span>Channel</span></a></li><li class="level-2"><a class="final" href="/datacollection/batchimport/"><span>Importing Data in Batch</span></a></li><li class="level-2"><a class="final" href="/datacollection/analytics/"><span>Using Analytics Tools</span></a></li></ul></li><li class="level-1"><a cl ass="expandible" href="#"><span>Choosing an Algorithm(s)</span></a><ul><li class="level-2"><a class="final" href="/algorithm/"><span>Built-in Algorithm Libraries</span></a></li><li class="level-2"><a class="final" href="/algorithm/switch/"><span>Switching to Another Algorithm</span></a></li><li class="level-2"><a class="final" href="/algorithm/multiple/"><span>Combining Multiple Algorithms</span></a></li><li class="level-2"><a class="final" href="/algorithm/custom/"><span>Adding Your Own Algorithms</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>ML Tuning and Evaluation</span></a><ul><li class="level-2"><a class="final" href="/evaluation/"><span>Overview</span></a></li><li class="level-2"><a class="final" href="/evaluation/paramtuning/"><span>Hyperparameter Tuning</span></a></li><li class="level-2"><a class="final" href="/evaluation/evaluationdashboard/"><span>Evaluation Dashboard</span></a></li><li class="level-2"><a class="final" href="/evaluation /metricchoose/"><span>Choosing Evaluation Metrics</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricbuild/"><span>Building Evaluation Metrics</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>System Architecture</span></a><ul><li class="level-2"><a class="final" href="/system/"><span>Architecture Overview</span></a></li><li class="level-2"><a class="final" href="/system/anotherdatastore/"><span>Using Another Data Store</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Engine Template Gallery</span></a><ul><li class="level-2"><a class="final" href="/gallery/template-gallery/"><span>Browse</span></a></li><li class="level-2"><a class="final" href="/community/submit-template/"><span>Submit your Engine as a Template</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Demo Tutorials</span></a><ul><li class="level-2"><a class="final" href="/demo/tapster/"><span>Comics Rec ommendation Demo</span></a></li><li class="level-2"><a class="final" href="/demo/community/"><span>Community Contributed Demo</span></a></li><li class="level-2"><a class="final" href="/demo/textclassification/"><span>Text Classification Engine Tutorial</span></a></li></ul></li><li class="level-1"><a class="expandible" href="/community/"><span>Getting Involved</span></a><ul><li class="level-2"><a class="final" href="/community/contribute-code/"><span>Contribute Code</span></a></li><li class="level-2"><a class="final" href="/community/contribute-documentation/"><span>Contribute Documentation</span></a></li><li class="level-2"><a class="final" href="/community/contribute-sdk/"><span>Contribute a SDK</span></a></li><li class="level-2"><a class="final" href="/community/contribute-webhook/"><span>Contribute a Webhook</span></a></li><li class="level-2"><a class="final" href="/community/projects/"><span>Community Projects</span></a></li></ul></li><li class="level-1"><a class="expandible" hr ef="#"><span>Getting Help</span></a><ul><li class="level-2"><a class="final" href="/resources/faq/"><span>FAQs</span></a></li><li class="level-2"><a class="final" href="/support/"><span>Support</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Resources</span></a><ul><li class="level-2"><a class="final" href="/resources/intellij/"><span>Developing Engines with IntelliJ IDEA</span></a></li><li class="level-2"><a class="final" href="/resources/upgrade/"><span>Upgrade Instructions</span></a></li><li class="level-2"><a class="final" href="/resources/glossary/"><span>Glossary</span></a></li></ul></li></ul></nav></div><div class="col-md-9 col-sm-12"><div class="content-header hidden-md hidden-lg"><div id="page-title"><h1>Dimensionality Reduction With PredictionIO</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#data-example">Data Example</a> </li> <li> <a href="#principal-component-ana lysis">Principal Component Analysis</a> </li> <li> <a href="#modifying-the-engine-template">Modifying the Engine Template</a> </li> <li> <a href="#testing-the-engine">Testing the Engine</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/tree/livedoc/docs/manual/source/machinelearning/dimensionalityreduction.html.md";><img src="/images/icons/edit-pencil-d6c1bb3d.png"/>Edit this page</a></div><div class="content-header hidden-sm hidden-xs"><div id="page-title"><h1>Dimensionality Reduction With PredictionIO</h1></div></div><div class="content"><p>The purpose of this guide is to teach developers how to incorporate "dimensionality reduction" into a PredictionIO engine <a href="https://en.wikipedia.org/wiki/Principal_component_analysis";>Principal Component Analysis</a> (PCA) on the <a href="https://www.kaggle.com/c/digit-recognizer";>MNIST digit recognition dataset</a>. To do this, you will be modifying the PredictionIO <a hr ef="http://templates.prediction.io/PredictionIO/template-scala-parallel-classification";>classification engine template</a>. This guide will demonstrate how to import the specific data set in batch, and also how to change the engine components in order to incorporate the new sample data and implement PCA.</p><p>In machine learning, specifically in <a href="http://en.wikipedia.org/wiki/Supervised_learning";>supervised learning</a>, the general problem at hand is to predict a numeric outcome \(y\) from a numeric vector \(\bf{x}\). The different components of \(\bf{x}\) are called <strong>features</strong>, and usually represent observed values such as a hospital patient's age, weight, height, sex, etc. There are subtle issues that begin to arise as the number of features contained in each feature vector increases. We briefly list some of the issues that arise as the number of features grows in size:</p> <ul> <li><p><strong>Computation:</strong> The time complexity of machine learnin g algorithms often times depends on the number of features used. That is, the more features one uses for prediction, the more time it takes to train a model.</p></li> <li><p><strong>Prediction Performance:</strong> Often times there will be features that, when used in training, will actually decrease the predictive performance of a particular algorithm. </p></li> <li><p><strong>Curse of Dimensionality:</strong> It is harder to make inference and predictions in high dimensional spaces simply due to the fact that we need to sample a lot more observations. Think about it in this way, suppose that we sample 100 points lying on a flat solid square, and 100 points in a solid cube. The 100 points from the square will likely take up a larger proportion of its area, in comparison to the proportion of the cube's volume that the points sampled from it occupy. Hence we would need to sample more points from the cube in order to get better estimates of the different properties of the cube, su ch as height, length, and width. This is shown in the following figure:</p></li> </ul> <table><thead> <tr> <th>100 Points Sampled From Unit Square</th> <th>100 Points Sampled From Unit Cube</th> </tr> </thead><tbody> <tr> <td></td> <td></td> </tr> <tr> <td><img alt="Square Samples" src="/images/machinelearning/featureselection/square100-df83c1ae.png"/></td> <td><img alt="Cube Samples" src="/images/machinelearning/featureselection/cube100-a8fe5433.png"/></td> </tr> <tr> <td></td> <td></td> </tr> </tbody></table> <p>Dimensionality reduction is the process of applying a transformation to your feature vectors in order to produce a vector with the same or less number of features. Principal component Analysis (PCA) is a technique for dimensionality reduction. This can be treated as a data processing technique, and so with respect to the <a href="/customize/">DASE</a> framework, it will fall into the Data Preparator engine component. </p><p>This guide will also help to solidify the concept of taking an engine template and customizing it for a particular use case: hand-written numeric digit recognition.</p><h2 id='data-example' class='header-anchors'>Data Example</h2><p>As a guiding example, a base data set, the <a href="https://www.kaggle.com/c/digit-recognizer/data";>MNIST digit recognition dataset</a>, is used. This is a perfect data set for dimensionality reduction, for, in this data set, the features that will be used for learning are pixel entries in a \(28 \times 28\) pixel image. There is really no direct interpretation of any one feature, so that you do not lose anything in applying a transformation that will treat the features as <a href="https://en.wikipedia.org/wiki/Linear_combination";>linear combinations</a> of some set "convenient" vectors. </p><p>Now, we first pull the <a href="http://templates.prediction.io/PredictionIO/template-scala-parallel-classification";>classification engine template</a> via the following bash line</p><div class="highlig ht shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>pio template get PredictionIO/template-scala-parallel-classification <Your new engine directory> +<!DOCTYPE html><html><head><title>Dimensionality Reduction With PredictionIO</title><meta charset="utf-8"/><meta content="IE=edge,chrome=1" http-equiv="X-UA-Compatible"/><meta name="viewport" content="width=device-width, initial-scale=1.0"/><meta class="swiftype" name="title" data-type="string" content="Dimensionality Reduction With PredictionIO"/><link rel="canonical" href="https://docs.prediction.io/machinelearning/dimensionalityreduction/"/><link href="/images/favicon/normal-b330020a.png" rel="shortcut icon"/><link href="/images/favicon/apple-c0febcf2.png" rel="apple-touch-icon"/><link href="//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet"/><link href="//maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet"/><link href="/stylesheets/application-a2a2f408.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min. js"></script><script src="//cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}catch(e){}</script></head><body><div id="global"><header><div class="container" id="header-wrapper"><div class="row"><div class="col-sm-12"><div id="logo-wrapper"><span id="drawer-toggle"></span><a href="#"></a><a href="http://predictionio.incubator.apache.org/";><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="/gallery/template-gallery">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-predictionio/">OPEN SOURCE</a></div></div><img class="mobile-search-bar-toggler hidden-md hidden-lg" src="/images/icons/search-glass-704bd4ff.png"/></div></div></div></header><div id="search-bar-row-wrapper"><div class="container-fluid" id="search-bar-row"><div class="row">< div class="col-md-9 col-sm-11 col-xs-11"><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>Dimensionality Reduction With PredictionIO</h4></div><h4 class="hidden-sm hidden-xs">PredictionIO Docs</h4></div><div class="col-md-3 col-sm-1 col-xs-1 hidden-md hidden-lg"><img id="left-menu-indicator" src="/images/icons/down-arrow-dfe9f7fe.png"/></div><div class="col-md-3 col-sm-12 col-xs-12 swiftype-wrapper"><div class="swiftype"><form class="search-form"><img class="search-box-toggler hidden-xs hidden-sm" src="/images/icons/search-glass-704bd4ff.png"/><div class="search-box"><img src="/images/icons/search-glass-704bd4ff.png"/><input type="text" id="st-search-input" class="st-search-input" placeholder="Search Doc..."/></div><img class="swiftype-row-hider hidden-md hidden-lg" src="/images/icons/drawer-toggle-active-fcbef12a.png"/></form></div></div><div class="mobile-left-menu-toggler hidden-md hidden-lg"></div></div></div></div><div id="page" clas s="container-fluid"><div class="row"><div id="left-menu-wrapper" class="col-md-3"><nav id="nav-main"><ul><li class="level-1"><a class="expandible" href="/"><span>Apache PredictionIO (incubating) Documentation</span></a><ul><li class="level-2"><a class="final" href="/"><span>Welcome to Apache PredictionIO (incubating)</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Started</span></a><ul><li class="level-2"><a class="final" href="/start/"><span>A Quick Intro</span></a></li><li class="level-2"><a class="final" href="/install/"><span>Installing Apache PredictionIO (incubating)</span></a></li><li class="level-2"><a class="final" href="/start/download/"><span>Downloading an Engine Template</span></a></li><li class="level-2"><a class="final" href="/start/deploy/"><span>Deploying Your First Engine</span></a></li><li class="level-2"><a class="final" href="/start/customize/"><span>Customizing the Engine</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Integrating with Your App</span></a><ul><li class="level-2"><a class="final" href="/appintegration/"><span>App Integration Overview</span></a></li><li class="level-2"><a class="expandible" href="/sdk/"><span>List of SDKs</span></a><ul><li class="level-3"><a class="final" href="/sdk/java/"><span>Java & Android SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/php/"><span>PHP SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/python/"><span>Python SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/ruby/"><span>Ruby SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/community/"><span>Community Powered SDKs</span></a></li></ul></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Deploying an Engine</span></a><ul><li class="level-2"><a class="final" href="/deploy/"><span>Deploying as a Web Service</span></a></li><li class="level-2"><a class="final" href="/cli/#engine-comm ands"><span>Engine Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</span></a></li><li class="level-2"><a class="final" href="/deploy/engineparams/"><span>Setting Engine Parameters</span></a></li><li class="level-2"><a class="final" href="/deploy/enginevariants/"><span>Deploying Multiple Engine Variants</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Customizing an Engine</span></a><ul><li class="level-2"><a class="final" href="/customize/"><span>Learning DASE</span></a></li><li class="level-2"><a class="final" href="/customize/dase/"><span>Implement DASE</span></a></li><li class="level-2"><a class="final" href="/customize/troubleshooting/"><span>Troubleshooting Engine Development</span></a></li><li class="level-2"><a class="final" href="/api/current/#package"><span>Engine Scala APIs</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Collectin g and Analyzing Data</span></a><ul><li class="level-2"><a class="final" href="/datacollection/"><span>Event Server Overview</span></a></li><li class="level-2"><a class="final" href="/cli/#event-server-commands"><span>Event Server Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventapi/"><span>Collecting Data with REST/SDKs</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventmodel/"><span>Events Modeling</span></a></li><li class="level-2"><a class="final" href="/datacollection/webhooks/"><span>Unifying Multichannel Data with Webhooks</span></a></li><li class="level-2"><a class="final" href="/datacollection/channel/"><span>Channel</span></a></li><li class="level-2"><a class="final" href="/datacollection/batchimport/"><span>Importing Data in Batch</span></a></li><li class="level-2"><a class="final" href="/datacollection/analytics/"><span>Using Analytics Tools</span></a></li></ul></li><li class="level-1"><a cla ss="expandible" href="#"><span>Choosing an Algorithm(s)</span></a><ul><li class="level-2"><a class="final" href="/algorithm/"><span>Built-in Algorithm Libraries</span></a></li><li class="level-2"><a class="final" href="/algorithm/switch/"><span>Switching to Another Algorithm</span></a></li><li class="level-2"><a class="final" href="/algorithm/multiple/"><span>Combining Multiple Algorithms</span></a></li><li class="level-2"><a class="final" href="/algorithm/custom/"><span>Adding Your Own Algorithms</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>ML Tuning and Evaluation</span></a><ul><li class="level-2"><a class="final" href="/evaluation/"><span>Overview</span></a></li><li class="level-2"><a class="final" href="/evaluation/paramtuning/"><span>Hyperparameter Tuning</span></a></li><li class="level-2"><a class="final" href="/evaluation/evaluationdashboard/"><span>Evaluation Dashboard</span></a></li><li class="level-2"><a class="final" href="/evaluation/ metricchoose/"><span>Choosing Evaluation Metrics</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricbuild/"><span>Building Evaluation Metrics</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>System Architecture</span></a><ul><li class="level-2"><a class="final" href="/system/"><span>Architecture Overview</span></a></li><li class="level-2"><a class="final" href="/system/anotherdatastore/"><span>Using Another Data Store</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Engine Template Gallery</span></a><ul><li class="level-2"><a class="final" href="/gallery/template-gallery/"><span>Browse</span></a></li><li class="level-2"><a class="final" href="/community/submit-template/"><span>Submit your Engine as a Template</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Demo Tutorials</span></a><ul><li class="level-2"><a class="final" href="/demo/tapster/"><span>Comics Reco mmendation Demo</span></a></li><li class="level-2"><a class="final" href="/demo/community/"><span>Community Contributed Demo</span></a></li><li class="level-2"><a class="final" href="/demo/textclassification/"><span>Text Classification Engine Tutorial</span></a></li></ul></li><li class="level-1"><a class="expandible" href="/community/"><span>Getting Involved</span></a><ul><li class="level-2"><a class="final" href="/community/contribute-code/"><span>Contribute Code</span></a></li><li class="level-2"><a class="final" href="/community/contribute-documentation/"><span>Contribute Documentation</span></a></li><li class="level-2"><a class="final" href="/community/contribute-sdk/"><span>Contribute a SDK</span></a></li><li class="level-2"><a class="final" href="/community/contribute-webhook/"><span>Contribute a Webhook</span></a></li><li class="level-2"><a class="final" href="/community/projects/"><span>Community Projects</span></a></li></ul></li><li class="level-1"><a class="expandible" hre f="#"><span>Getting Help</span></a><ul><li class="level-2"><a class="final" href="/resources/faq/"><span>FAQs</span></a></li><li class="level-2"><a class="final" href="/support/"><span>Support</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Resources</span></a><ul><li class="level-2"><a class="final" href="/resources/intellij/"><span>Developing Engines with IntelliJ IDEA</span></a></li><li class="level-2"><a class="final" href="/resources/upgrade/"><span>Upgrade Instructions</span></a></li><li class="level-2"><a class="final" href="/resources/glossary/"><span>Glossary</span></a></li></ul></li></ul></nav></div><div class="col-md-9 col-sm-12"><div class="content-header hidden-md hidden-lg"><div id="page-title"><h1>Dimensionality Reduction With PredictionIO</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#data-example">Data Example</a> </li> <li> <a href="#principal-component-anal ysis">Principal Component Analysis</a> </li> <li> <a href="#modifying-the-engine-template">Modifying the Engine Template</a> </li> <li> <a href="#testing-the-engine">Testing the Engine</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/tree/livedoc/docs/manual/source/machinelearning/dimensionalityreduction.html.md";><img src="/images/icons/edit-pencil-d6c1bb3d.png"/>Edit this page</a></div><div class="content-header hidden-sm hidden-xs"><div id="page-title"><h1>Dimensionality Reduction With PredictionIO</h1></div></div><div class="content"><p>The purpose of this guide is to teach developers how to incorporate "dimensionality reduction" into a PredictionIO engine <a href="https://en.wikipedia.org/wiki/Principal_component_analysis";>Principal Component Analysis</a> (PCA) on the <a href="https://www.kaggle.com/c/digit-recognizer";>MNIST digit recognition dataset</a>. To do this, you will be modifying the PredictionIO <a hre f="/gallery/template-gallery/#classification">classification engine template</a>. This guide will demonstrate how to import the specific data set in batch, and also how to change the engine components in order to incorporate the new sample data and implement PCA.</p><p>In machine learning, specifically in <a href="http://en.wikipedia.org/wiki/Supervised_learning";>supervised learning</a>, the general problem at hand is to predict a numeric outcome \(y\) from a numeric vector \(\bf{x}\). The different components of \(\bf{x}\) are called <strong>features</strong>, and usually represent observed values such as a hospital patient's age, weight, height, sex, etc. There are subtle issues that begin to arise as the number of features contained in each feature vector increases. We briefly list some of the issues that arise as the number of features grows in size:</p> <ul> <li><p><strong>Computation:</strong> The time complexity of machine learning algorithms often times depends on the nu mber of features used. That is, the more features one uses for prediction, the more time it takes to train a model.</p></li> <li><p><strong>Prediction Performance:</strong> Often times there will be features that, when used in training, will actually decrease the predictive performance of a particular algorithm. </p></li> <li><p><strong>Curse of Dimensionality:</strong> It is harder to make inference and predictions in high dimensional spaces simply due to the fact that we need to sample a lot more observations. Think about it in this way, suppose that we sample 100 points lying on a flat solid square, and 100 points in a solid cube. The 100 points from the square will likely take up a larger proportion of its area, in comparison to the proportion of the cube's volume that the points sampled from it occupy. Hence we would need to sample more points from the cube in order to get better estimates of the different properties of the cube, such as height, length, and width. This is s hown in the following figure:</p></li> </ul> <table><thead> <tr> <th>100 Points Sampled From Unit Square</th> <th>100 Points Sampled From Unit Cube</th> </tr> </thead><tbody> <tr> <td></td> <td></td> </tr> <tr> <td><img alt="Square Samples" src="/images/machinelearning/featureselection/square100-df83c1ae.png"/></td> <td><img alt="Cube Samples" src="/images/machinelearning/featureselection/cube100-a8fe5433.png"/></td> </tr> <tr> <td></td> <td></td> </tr> </tbody></table> <p>Dimensionality reduction is the process of applying a transformation to your feature vectors in order to produce a vector with the same or less number of features. Principal component Analysis (PCA) is a technique for dimensionality reduction. This can be treated as a data processing technique, and so with respect to the <a href="/customize/">DASE</a> framework, it will fall into the Data Preparator engine component. </p><p>This guide will also help to solidify the concept of taking an engine template and customiz ing it for a particular use case: hand-written numeric digit recognition.</p><h2 id='data-example' class='header-anchors'>Data Example</h2><p>As a guiding example, a base data set, the <a href="https://www.kaggle.com/c/digit-recognizer/data";>MNIST digit recognition dataset</a>, is used. This is a perfect data set for dimensionality reduction, for, in this data set, the features that will be used for learning are pixel entries in a \(28 \times 28\) pixel image. There is really no direct interpretation of any one feature, so that you do not lose anything in applying a transformation that will treat the features as <a href="https://en.wikipedia.org/wiki/Linear_combination";>linear combinations</a> of some set "convenient" vectors. </p><p>Now, we first pull the <a href="/gallery/template-gallery/#classification">classification engine template</a> via the following bash line</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style=" text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>pio template get PredictionIO/template-scala-parallel-classification <Your new engine directory> </pre></td></tr></tbody></table> </div> <p>You should immediately be prompted with the following message:</p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>Please enter the template<span class="s1">'s Scala package name (e.g. com.mycompany): </span></pre></td></tr></tbody></table> </div> <p>Go ahead and input <code>FeatureReduction</code>, and feel free to just press enter for the remaining message prompts. For the remainder of this guide, you will be working in your new engine directory, so go ahead and <code>cd</code> into your new engine directory. At this point, go ahead and run the command </p><div class="highlight shell"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1</pre></td><td class="code"><pre>pio build </pre></td></tr></tbody></table> </div> <p>This will make sure that the PredictionIO dependency version for your project matches the version installed on your computer. Now, download the MNIST <code>train.csv</code> data set from the link above, and put this file in the <code>data</code> directory contained in the new engine directory. </p><h3 id='<strong>optional</strong>:-visualizing-observations' class='header-anchors' ><strong>Optional</strong>: Visualizing Observations</h3><p>If you want to actually convert the observation pixel data to an image go ahead and create a Python script called <code>picture_processing.py</code> into your data directory and copy and paste the following code into the script:</p><div class="highlight python"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right"><pre class="lineno">1 @@ -354,7 +354,7 @@ Import digit recognition data. <span class="k">new</span> <span class="nc">PreparedData</span><span class="o">(</span><span class="n">trainingData</span><span class="o">.</span><span class="n">observations</span><span class="o">,</span> <span class="n">pp</span><span class="o">)</span> <span class="o">}</span> <span class="o">}</span> -</pre></td></tr></tbody></table> </div> <p>The Data Preparator engine component is now complete, and we can move on to the Algorithm component.</p><h3 id='algorithm-modifications' class='header-anchors'>Algorithm Modifications</h3><p>The default algorithm used in the classification template is Naive Bayes. Now, this is a <a href="https://en.wikipedia.org/wiki/Probabilistic_classification";>probabilistic classifier</a> that makes certain assumptions about the data that do not really match the format of the PCA-transformed data. In particular, it assumes that the vectors consist of counts. In particular, this means it assumes non-negative feature values. However, upon applying PCA on the data, you have no guarantees that you will have purely non-negative features. Given this, you will delete the script <code>NaiveBayesAlgorithm.scala</code>, and create one called <code>LRAlgorithm.scala</code> (in the <code>src/main/scala/</code> directory) which implements <a href="https://en.wikipedi a.org/wiki/Multinomial_logistic_regression">Multinomial Logistic Regression</a>. </p><p>The implementation details are not discussed in this guide, as the point of this guide is to show how to incorporate <strong>dimensionality reduction</strong> techniques by incorporating PCA. The latter paragraph is mentioned in order to emphasize the fact that applying the PCA transformation (or possibly other dimensionality reduction techniques) will largely remove the interpretability of features, so that model assumptions relying on such interpretations may no longer be satisfied. This is just something to keep in mind.</p><p>The following code is taken from the <a href="http://templates.prediction.io/PredictionIO/template-scala-parallel-textclassification";>text classification engine template</a> and adapted to match the project definitions. Copy and paste into the new scala script, <code>LRAlgorithm.scala</code>: </p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><t d class="gutter gl" style="text-align: right"><pre class="lineno">1 +</pre></td></tr></tbody></table> </div> <p>The Data Preparator engine component is now complete, and we can move on to the Algorithm component.</p><h3 id='algorithm-modifications' class='header-anchors'>Algorithm Modifications</h3><p>The default algorithm used in the classification template is Naive Bayes. Now, this is a <a href="https://en.wikipedia.org/wiki/Probabilistic_classification";>probabilistic classifier</a> that makes certain assumptions about the data that do not really match the format of the PCA-transformed data. In particular, it assumes that the vectors consist of counts. In particular, this means it assumes non-negative feature values. However, upon applying PCA on the data, you have no guarantees that you will have purely non-negative features. Given this, you will delete the script <code>NaiveBayesAlgorithm.scala</code>, and create one called <code>LRAlgorithm.scala</code> (in the <code>src/main/scala/</code> directory) which implements <a href="https://en.wikipedi a.org/wiki/Multinomial_logistic_regression">Multinomial Logistic Regression</a>. </p><p>The implementation details are not discussed in this guide, as the point of this guide is to show how to incorporate <strong>dimensionality reduction</strong> techniques by incorporating PCA. The latter paragraph is mentioned in order to emphasize the fact that applying the PCA transformation (or possibly other dimensionality reduction techniques) will largely remove the interpretability of features, so that model assumptions relying on such interpretations may no longer be satisfied. This is just something to keep in mind.</p><p>The following code is taken from the <a href="/gallery/template-gallery/#classification">text classification engine template</a> and adapted to match the project definitions. Copy and paste into the new scala script, <code>LRAlgorithm.scala</code>: </p><div class="highlight scala"><table style="border-spacing: 0"><tbody><tr><td class="gutter gl" style="text-align: right" ><pre class="lineno">1 2 3 4 http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/machinelearning/dimensionalityreduction/index.html.gz ---------------------------------------------------------------------- diff --git a/machinelearning/dimensionalityreduction/index.html.gz b/machinelearning/dimensionalityreduction/index.html.gz index bec8784..dc77327 100644 Binary files a/machinelearning/dimensionalityreduction/index.html.gz and b/machinelearning/dimensionalityreduction/index.html.gz differ http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/machinelearning/modelingworkflow/index.html ---------------------------------------------------------------------- diff --git a/machinelearning/modelingworkflow/index.html b/machinelearning/modelingworkflow/index.html index 746f3c8..dba5d0e 100644 --- a/machinelearning/modelingworkflow/index.html +++ b/machinelearning/modelingworkflow/index.html @@ -1,4 +1,4 @@ -<!DOCTYPE html><html><head><title>Modeling Workflow and DASE</title><meta charset="utf-8"/><meta content="IE=edge,chrome=1" http-equiv="X-UA-Compatible"/><meta name="viewport" content="width=device-width, initial-scale=1.0"/><meta class="swiftype" name="title" data-type="string" content="Modeling Workflow and DASE"/><link rel="canonical" href="https://docs.prediction.io/machinelearning/modelingworkflow/"/><link href="/images/favicon/normal-b330020a.png" rel="shortcut icon"/><link href="/images/favicon/apple-c0febcf2.png" rel="apple-touch-icon"/><link href="//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet"/><link href="//maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet"/><link href="/stylesheets/application-a2a2f408.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min.js"></script><script src="//cdn.mathjax .org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}catch(e){}</script></head><body><div id="global"><header><div class="container" id="header-wrapper"><div class="row"><div class="col-sm-12"><div id="logo-wrapper"><span id="drawer-toggle"></span><a href="#"></a><a href="http://predictionio.incubator.apache.org/";><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="//templates.prediction.io/">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-predictionio/">OPEN SOURCE</a></div></div><img class="mobile-search-bar-toggler hidden-md hidden-lg" src="/images/icons/search-glass-704bd4ff.png"/></div></div></div></header><div id="search-bar-row-wrapper"><div class="container-fluid" id="search-bar-row"><div class="row"><div class="col-md-9 col-sm-11 col-xs-1 1"><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>Modeling Workflow and DASE</h4></div><h4 class="hidden-sm hidden-xs">PredictionIO Docs</h4></div><div class="col-md-3 col-sm-1 col-xs-1 hidden-md hidden-lg"><img id="left-menu-indicator" src="/images/icons/down-arrow-dfe9f7fe.png"/></div><div class="col-md-3 col-sm-12 col-xs-12 swiftype-wrapper"><div class="swiftype"><form class="search-form"><img class="search-box-toggler hidden-xs hidden-sm" src="/images/icons/search-glass-704bd4ff.png"/><div class="search-box"><img src="/images/icons/search-glass-704bd4ff.png"/><input type="text" id="st-search-input" class="st-search-input" placeholder="Search Doc..."/></div><img class="swiftype-row-hider hidden-md hidden-lg" src="/images/icons/drawer-toggle-active-fcbef12a.png"/></form></div></div><div class="mobile-left-menu-toggler hidden-md hidden-lg"></div></div></div></div><div id="page" class="container-fluid"><div class="row"><div id="left-men u-wrapper" class="col-md-3"><nav id="nav-main"><ul><li class="level-1"><a class="expandible" href="/"><span>Apache PredictionIO (incubating) Documentation</span></a><ul><li class="level-2"><a class="final" href="/"><span>Welcome to Apache PredictionIO (incubating)</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Started</span></a><ul><li class="level-2"><a class="final" href="/start/"><span>A Quick Intro</span></a></li><li class="level-2"><a class="final" href="/install/"><span>Installing Apache PredictionIO (incubating)</span></a></li><li class="level-2"><a class="final" href="/start/download/"><span>Downloading an Engine Template</span></a></li><li class="level-2"><a class="final" href="/start/deploy/"><span>Deploying Your First Engine</span></a></li><li class="level-2"><a class="final" href="/start/customize/"><span>Customizing the Engine</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Integrating with You r App</span></a><ul><li class="level-2"><a class="final" href="/appintegration/"><span>App Integration Overview</span></a></li><li class="level-2"><a class="expandible" href="/sdk/"><span>List of SDKs</span></a><ul><li class="level-3"><a class="final" href="/sdk/java/"><span>Java & Android SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/php/"><span>PHP SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/python/"><span>Python SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/ruby/"><span>Ruby SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/community/"><span>Community Powered SDKs</span></a></li></ul></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Deploying an Engine</span></a><ul><li class="level-2"><a class="final" href="/deploy/"><span>Deploying as a Web Service</span></a></li><li class="level-2"><a class="final" href="/cli/#engine-commands"><span>Engine Command-line Interface</span></a></ li><li class="level-2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</span></a></li><li class="level-2"><a class="final" href="/deploy/engineparams/"><span>Setting Engine Parameters</span></a></li><li class="level-2"><a class="final" href="/deploy/enginevariants/"><span>Deploying Multiple Engine Variants</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Customizing an Engine</span></a><ul><li class="level-2"><a class="final" href="/customize/"><span>Learning DASE</span></a></li><li class="level-2"><a class="final" href="/customize/dase/"><span>Implement DASE</span></a></li><li class="level-2"><a class="final" href="/customize/troubleshooting/"><span>Troubleshooting Engine Development</span></a></li><li class="level-2"><a class="final" href="/api/current/#package"><span>Engine Scala APIs</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Collecting and Analyzing Data</span></a><ul><li class="level-2" ><a class="final" href="/datacollection/"><span>Event Server >Overview</span></a></li><li class="level-2"><a class="final" >href="/cli/#event-server-commands"><span>Event Server Command-line >Interface</span></a></li><li class="level-2"><a class="final" >href="/datacollection/eventapi/"><span>Collecting Data with >REST/SDKs</span></a></li><li class="level-2"><a class="final" >href="/datacollection/eventmodel/"><span>Events Modeling</span></a></li><li >class="level-2"><a class="final" >href="/datacollection/webhooks/"><span>Unifying Multichannel Data with >Webhooks</span></a></li><li class="level-2"><a class="final" >href="/datacollection/channel/"><span>Channel</span></a></li><li >class="level-2"><a class="final" >href="/datacollection/batchimport/"><span>Importing Data in >Batch</span></a></li><li class="level-2"><a class="final" >href="/datacollection/analytics/"><span>Using Analytics >Tools</span></a></li></ul></li><li class="level-1"><a class="expandible" >href="#"><span>Choosing an Algorithm(s )</span></a><ul><li class="level-2"><a class="final" href="/algorithm/"><span>Built-in Algorithm Libraries</span></a></li><li class="level-2"><a class="final" href="/algorithm/switch/"><span>Switching to Another Algorithm</span></a></li><li class="level-2"><a class="final" href="/algorithm/multiple/"><span>Combining Multiple Algorithms</span></a></li><li class="level-2"><a class="final" href="/algorithm/custom/"><span>Adding Your Own Algorithms</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>ML Tuning and Evaluation</span></a><ul><li class="level-2"><a class="final" href="/evaluation/"><span>Overview</span></a></li><li class="level-2"><a class="final" href="/evaluation/paramtuning/"><span>Hyperparameter Tuning</span></a></li><li class="level-2"><a class="final" href="/evaluation/evaluationdashboard/"><span>Evaluation Dashboard</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricchoose/"><span>Choosing Evaluation Metrics</span ></a></li><li class="level-2"><a class="final" >href="/evaluation/metricbuild/"><span>Building Evaluation >Metrics</span></a></li></ul></li><li class="level-1"><a class="expandible" >href="#"><span>System Architecture</span></a><ul><li class="level-2"><a >class="final" href="/system/"><span>Architecture Overview</span></a></li><li >class="level-2"><a class="final" href="/system/anotherdatastore/"><span>Using >Another Data Store</span></a></li></ul></li><li class="level-1"><a >class="expandible" href="#"><span>Engine Template Gallery</span></a><ul><li >class="level-2"><a class="final" >href="/gallery/template-gallery/"><span>Browse</span></a></li><li >class="level-2"><a class="final" >href="/community/submit-template/"><span>Submit your Engine as a >Template</span></a></li></ul></li><li class="level-1"><a class="expandible" >href="#"><span>Demo Tutorials</span></a><ul><li class="level-2"><a >class="final" href="/demo/tapster/"><span>Comics Recommendation >Demo</span></a></li><li class="level-2"><a class="final" href="/demo/community/"><span>Community Contributed Demo</span></a></li><li class="level-2"><a class="final" href="/demo/textclassification/"><span>Text Classification Engine Tutorial</span></a></li></ul></li><li class="level-1"><a class="expandible" href="/community/"><span>Getting Involved</span></a><ul><li class="level-2"><a class="final" href="/community/contribute-code/"><span>Contribute Code</span></a></li><li class="level-2"><a class="final" href="/community/contribute-documentation/"><span>Contribute Documentation</span></a></li><li class="level-2"><a class="final" href="/community/contribute-sdk/"><span>Contribute a SDK</span></a></li><li class="level-2"><a class="final" href="/community/contribute-webhook/"><span>Contribute a Webhook</span></a></li><li class="level-2"><a class="final" href="/community/projects/"><span>Community Projects</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Help</span></a><ul><li class="leve l-2"><a class="final" href="/resources/faq/"><span>FAQs</span></a></li><li class="level-2"><a class="final" href="/support/"><span>Support</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Resources</span></a><ul><li class="level-2"><a class="final" href="/resources/intellij/"><span>Developing Engines with IntelliJ IDEA</span></a></li><li class="level-2"><a class="final" href="/resources/upgrade/"><span>Upgrade Instructions</span></a></li><li class="level-2"><a class="final" href="/resources/glossary/"><span>Glossary</span></a></li></ul></li></ul></nav></div><div class="col-md-9 col-sm-12"><div class="content-header hidden-md hidden-lg"><div id="page-title"><h1>Modeling Workflow and DASE</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#training-the-model">Training The Model</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/ tree/livedoc/docs/manual/source/machinelearning/modelingworkflow.html.md"><img src="/images/icons/edit-pencil-d6c1bb3d.png"/>Edit this page</a></div><div class="content-header hidden-sm hidden-xs"><div id="page-title"><h1>Modeling Workflow and DASE</h1></div></div><div class="content"><p>In addition to the DASE components, we also introduce the Data Model and Training Model abstractions. The Data Model abstraction refers to the set of Scala classes dealing with the implementation of modeling choices relating to feature <strong>extraction</strong>, <strong>preparation</strong>, and/or <strong>selection</strong>. For this illustration, this only includes the vectorization of text and t.f.-i.d.f. processing which is entirely implemented in the PreparedData class. The Training Model abstraction refers to any set of classes that individually take in a set of feature observations and output a predictive model. This predictive model is leveraged by the Algorithm component to produce predic tion results to queries in real-time. In the engine template, this abstraction is implemented in the NBModel class. <strong>Please note that these are conceptual abstractions that are designed to make engine development easier by decoupling class functionality.</strong> Keeping these abstractions in mind will help you in the future with debugging your code, and also make it easier to incorporate different modeling ideas into your engine. </p><p>The figure below shows a graphical representation of the engine architecture just described, as well as its interactions with your web/app and a provided Event Server:</p><p><img alt="Engine Overview" src="/images/demo/text_classification_template/engine_overview-27e09a89.png"/></p><h2 id='training-the-model' class='header-anchors'>Training The Model</h2><p>This section will guide you through the two Training Model implementations that come with this engine template. Recall that the Training Model abstraction refers to an arbitrary set Scala Class that outputs a predictive model (i.e. implements some method that can be used for prediction). The general problem this engine template is tackling is text classification, so that our Training Model abstraction domain is restricted to implementations producing classifiers. In particular, the classification model that is implemented in this engine template is based on Multinomial Naive Bayes using t.f.-i.d.f. vectorized text. </p></div></div></div></div><footer><div class="container"><div class="seperator"></div><div class="row"><div class="col-md-6 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Community</h4><ul><li><a href="//docs.prediction.io/install/" target="blank">Download</a></li><li><a href="//docs.prediction.io/" target="blank">Docs</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">GitHub</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to User Mailing List</a></l i><li><a href="//stackoverflow.com/questions/tagged/predictionio" target="blank">Stackoverflow</a></li></ul></div></div><div class="col-md-6 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Contribute</h4><ul><li><a href="//predictionio.incubator.apache.org/community/contribute-code/" target="blank">Contribute</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">Source Code</a></li><li><a href="//issues.apache.org/jira/browse/PIO" target="blank">Bug Tracker</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to Development Mailing List</a></li></ul></div></div></div></div><div id="footer-bottom"><div class="container"><div class="row"><div class="col-md-12"><div id="footer-logo-wrapper"><img alt="PredictionIO" src="/images/logos/logo-white-d1e9c6e6.png"/></div><div id="social-icons-wrapper"><a class="github-button" href="https://github.com/apache/incubator-predictionio"; data-style="mega" data-count-href="/apache/incubator-predictionio/stargazers" data-count-api="/repos/apache/incubator-predictionio#stargazers_count" data-count-aria-label="# stargazers on GitHub" aria-label="Star apache/incubator-predictionio on GitHub">Star</a> <a class="github-button" href="https://github.com/apache/incubator-predictionio/fork"; data-icon="octicon-git-branch" data-style="mega" data-count-href="/apache/incubator-predictionio/network" data-count-api="/repos/apache/incubator-predictionio#forks_count" data-count-aria-label="# forks on GitHub" aria-label="Fork apache/incubator-predictionio on GitHub">Fork</a> <script id="github-bjs" async="" defer="" src="https://buttons.github.io/buttons.js";></script><a href="//www.facebook.com/predictionio" target="blank"><img alt="PredictionIO on Twitter" src="/images/icons/twitter-ea9dc152.png"/></a> <a href="//twitter.com/predictionio" target="blank"><img alt="PredictionIO on Facebook" src="/images/icons/facebook-5c57939c.png"/></a> </div></div></d iv></div></div></footer></div><script>(function(w,d,t,u,n,s,e){w['SwiftypeObject']=n;w[n]=w[n]||function(){ +<!DOCTYPE html><html><head><title>Modeling Workflow and DASE</title><meta charset="utf-8"/><meta content="IE=edge,chrome=1" http-equiv="X-UA-Compatible"/><meta name="viewport" content="width=device-width, initial-scale=1.0"/><meta class="swiftype" name="title" data-type="string" content="Modeling Workflow and DASE"/><link rel="canonical" href="https://docs.prediction.io/machinelearning/modelingworkflow/"/><link href="/images/favicon/normal-b330020a.png" rel="shortcut icon"/><link href="/images/favicon/apple-c0febcf2.png" rel="apple-touch-icon"/><link href="//fonts.googleapis.com/css?family=Open+Sans:300italic,400italic,600italic,700italic,800italic,400,300,600,700,800" rel="stylesheet"/><link href="//maxcdn.bootstrapcdn.com/font-awesome/4.2.0/css/font-awesome.min.css" rel="stylesheet"/><link href="/stylesheets/application-a2a2f408.css" rel="stylesheet" type="text/css"/><script src="//cdnjs.cloudflare.com/ajax/libs/html5shiv/3.7.2/html5shiv.min.js"></script><script src="//cdn.mathjax .org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script src="//use.typekit.net/pqo0itb.js"></script><script>try{Typekit.load({ async: true });}catch(e){}</script></head><body><div id="global"><header><div class="container" id="header-wrapper"><div class="row"><div class="col-sm-12"><div id="logo-wrapper"><span id="drawer-toggle"></span><a href="#"></a><a href="http://predictionio.incubator.apache.org/";><img alt="PredictionIO" id="logo" src="/images/logos/logo-ee2b9bb3.png"/></a></div><div id="menu-wrapper"><div id="pill-wrapper"><a class="pill left" href="/gallery/template-gallery">TEMPLATES</a> <a class="pill right" href="//github.com/apache/incubator-predictionio/">OPEN SOURCE</a></div></div><img class="mobile-search-bar-toggler hidden-md hidden-lg" src="/images/icons/search-glass-704bd4ff.png"/></div></div></div></header><div id="search-bar-row-wrapper"><div class="container-fluid" id="search-bar-row"><div class="row"><div class="col-md-9 col-sm-11 col-xs-11 "><div class="hidden-md hidden-lg" id="mobile-page-heading-wrapper"><p>PredictionIO Docs</p><h4>Modeling Workflow and DASE</h4></div><h4 class="hidden-sm hidden-xs">PredictionIO Docs</h4></div><div class="col-md-3 col-sm-1 col-xs-1 hidden-md hidden-lg"><img id="left-menu-indicator" src="/images/icons/down-arrow-dfe9f7fe.png"/></div><div class="col-md-3 col-sm-12 col-xs-12 swiftype-wrapper"><div class="swiftype"><form class="search-form"><img class="search-box-toggler hidden-xs hidden-sm" src="/images/icons/search-glass-704bd4ff.png"/><div class="search-box"><img src="/images/icons/search-glass-704bd4ff.png"/><input type="text" id="st-search-input" class="st-search-input" placeholder="Search Doc..."/></div><img class="swiftype-row-hider hidden-md hidden-lg" src="/images/icons/drawer-toggle-active-fcbef12a.png"/></form></div></div><div class="mobile-left-menu-toggler hidden-md hidden-lg"></div></div></div></div><div id="page" class="container-fluid"><div class="row"><div id="left-menu -wrapper" class="col-md-3"><nav id="nav-main"><ul><li class="level-1"><a class="expandible" href="/"><span>Apache PredictionIO (incubating) Documentation</span></a><ul><li class="level-2"><a class="final" href="/"><span>Welcome to Apache PredictionIO (incubating)</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Started</span></a><ul><li class="level-2"><a class="final" href="/start/"><span>A Quick Intro</span></a></li><li class="level-2"><a class="final" href="/install/"><span>Installing Apache PredictionIO (incubating)</span></a></li><li class="level-2"><a class="final" href="/start/download/"><span>Downloading an Engine Template</span></a></li><li class="level-2"><a class="final" href="/start/deploy/"><span>Deploying Your First Engine</span></a></li><li class="level-2"><a class="final" href="/start/customize/"><span>Customizing the Engine</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Integrating with Your App</span></a><ul><li class="level-2"><a class="final" href="/appintegration/"><span>App Integration Overview</span></a></li><li class="level-2"><a class="expandible" href="/sdk/"><span>List of SDKs</span></a><ul><li class="level-3"><a class="final" href="/sdk/java/"><span>Java & Android SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/php/"><span>PHP SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/python/"><span>Python SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/ruby/"><span>Ruby SDK</span></a></li><li class="level-3"><a class="final" href="/sdk/community/"><span>Community Powered SDKs</span></a></li></ul></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Deploying an Engine</span></a><ul><li class="level-2"><a class="final" href="/deploy/"><span>Deploying as a Web Service</span></a></li><li class="level-2"><a class="final" href="/cli/#engine-commands"><span>Engine Command-line Interface</span></a></l i><li class="level-2"><a class="final" href="/deploy/monitoring/"><span>Monitoring Engine</span></a></li><li class="level-2"><a class="final" href="/deploy/engineparams/"><span>Setting Engine Parameters</span></a></li><li class="level-2"><a class="final" href="/deploy/enginevariants/"><span>Deploying Multiple Engine Variants</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Customizing an Engine</span></a><ul><li class="level-2"><a class="final" href="/customize/"><span>Learning DASE</span></a></li><li class="level-2"><a class="final" href="/customize/dase/"><span>Implement DASE</span></a></li><li class="level-2"><a class="final" href="/customize/troubleshooting/"><span>Troubleshooting Engine Development</span></a></li><li class="level-2"><a class="final" href="/api/current/#package"><span>Engine Scala APIs</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Collecting and Analyzing Data</span></a><ul><li class="level-2"> <a class="final" href="/datacollection/"><span>Event Server Overview</span></a></li><li class="level-2"><a class="final" href="/cli/#event-server-commands"><span>Event Server Command-line Interface</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventapi/"><span>Collecting Data with REST/SDKs</span></a></li><li class="level-2"><a class="final" href="/datacollection/eventmodel/"><span>Events Modeling</span></a></li><li class="level-2"><a class="final" href="/datacollection/webhooks/"><span>Unifying Multichannel Data with Webhooks</span></a></li><li class="level-2"><a class="final" href="/datacollection/channel/"><span>Channel</span></a></li><li class="level-2"><a class="final" href="/datacollection/batchimport/"><span>Importing Data in Batch</span></a></li><li class="level-2"><a class="final" href="/datacollection/analytics/"><span>Using Analytics Tools</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Choosing an Algorithm(s) </span></a><ul><li class="level-2"><a class="final" href="/algorithm/"><span>Built-in Algorithm Libraries</span></a></li><li class="level-2"><a class="final" href="/algorithm/switch/"><span>Switching to Another Algorithm</span></a></li><li class="level-2"><a class="final" href="/algorithm/multiple/"><span>Combining Multiple Algorithms</span></a></li><li class="level-2"><a class="final" href="/algorithm/custom/"><span>Adding Your Own Algorithms</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>ML Tuning and Evaluation</span></a><ul><li class="level-2"><a class="final" href="/evaluation/"><span>Overview</span></a></li><li class="level-2"><a class="final" href="/evaluation/paramtuning/"><span>Hyperparameter Tuning</span></a></li><li class="level-2"><a class="final" href="/evaluation/evaluationdashboard/"><span>Evaluation Dashboard</span></a></li><li class="level-2"><a class="final" href="/evaluation/metricchoose/"><span>Choosing Evaluation Metrics</span> </a></li><li class="level-2"><a class="final" href="/evaluation/metricbuild/"><span>Building Evaluation Metrics</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>System Architecture</span></a><ul><li class="level-2"><a class="final" href="/system/"><span>Architecture Overview</span></a></li><li class="level-2"><a class="final" href="/system/anotherdatastore/"><span>Using Another Data Store</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Engine Template Gallery</span></a><ul><li class="level-2"><a class="final" href="/gallery/template-gallery/"><span>Browse</span></a></li><li class="level-2"><a class="final" href="/community/submit-template/"><span>Submit your Engine as a Template</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Demo Tutorials</span></a><ul><li class="level-2"><a class="final" href="/demo/tapster/"><span>Comics Recommendation Demo</span></a></li><li class="level-2"><a c lass="final" href="/demo/community/"><span>Community Contributed Demo</span></a></li><li class="level-2"><a class="final" href="/demo/textclassification/"><span>Text Classification Engine Tutorial</span></a></li></ul></li><li class="level-1"><a class="expandible" href="/community/"><span>Getting Involved</span></a><ul><li class="level-2"><a class="final" href="/community/contribute-code/"><span>Contribute Code</span></a></li><li class="level-2"><a class="final" href="/community/contribute-documentation/"><span>Contribute Documentation</span></a></li><li class="level-2"><a class="final" href="/community/contribute-sdk/"><span>Contribute a SDK</span></a></li><li class="level-2"><a class="final" href="/community/contribute-webhook/"><span>Contribute a Webhook</span></a></li><li class="level-2"><a class="final" href="/community/projects/"><span>Community Projects</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Getting Help</span></a><ul><li class="level -2"><a class="final" href="/resources/faq/"><span>FAQs</span></a></li><li class="level-2"><a class="final" href="/support/"><span>Support</span></a></li></ul></li><li class="level-1"><a class="expandible" href="#"><span>Resources</span></a><ul><li class="level-2"><a class="final" href="/resources/intellij/"><span>Developing Engines with IntelliJ IDEA</span></a></li><li class="level-2"><a class="final" href="/resources/upgrade/"><span>Upgrade Instructions</span></a></li><li class="level-2"><a class="final" href="/resources/glossary/"><span>Glossary</span></a></li></ul></li></ul></nav></div><div class="col-md-9 col-sm-12"><div class="content-header hidden-md hidden-lg"><div id="page-title"><h1>Modeling Workflow and DASE</h1></div></div><div id="table-of-content-wrapper"><h5>On this page</h5><aside id="table-of-contents"><ul> <li> <a href="#training-the-model">Training The Model</a> </li> </ul> </aside><hr/><a id="edit-page-link" href="https://github.com/apache/incubator-predictionio/t ree/livedoc/docs/manual/source/machinelearning/modelingworkflow.html.md"><img src="/images/icons/edit-pencil-d6c1bb3d.png"/>Edit this page</a></div><div class="content-header hidden-sm hidden-xs"><div id="page-title"><h1>Modeling Workflow and DASE</h1></div></div><div class="content"><p>In addition to the DASE components, we also introduce the Data Model and Training Model abstractions. The Data Model abstraction refers to the set of Scala classes dealing with the implementation of modeling choices relating to feature <strong>extraction</strong>, <strong>preparation</strong>, and/or <strong>selection</strong>. For this illustration, this only includes the vectorization of text and t.f.-i.d.f. processing which is entirely implemented in the PreparedData class. The Training Model abstraction refers to any set of classes that individually take in a set of feature observations and output a predictive model. This predictive model is leveraged by the Algorithm component to produce predict ion results to queries in real-time. In the engine template, this abstraction is implemented in the NBModel class. <strong>Please note that these are conceptual abstractions that are designed to make engine development easier by decoupling class functionality.</strong> Keeping these abstractions in mind will help you in the future with debugging your code, and also make it easier to incorporate different modeling ideas into your engine. </p><p>The figure below shows a graphical representation of the engine architecture just described, as well as its interactions with your web/app and a provided Event Server:</p><p><img alt="Engine Overview" src="/images/demo/text_classification_template/engine_overview-27e09a89.png"/></p><h2 id='training-the-model' class='header-anchors'>Training The Model</h2><p>This section will guide you through the two Training Model implementations that come with this engine template. Recall that the Training Model abstraction refers to an arbitrary set Scala C lass that outputs a predictive model (i.e. implements some method that can be used for prediction). The general problem this engine template is tackling is text classification, so that our Training Model abstraction domain is restricted to implementations producing classifiers. In particular, the classification model that is implemented in this engine template is based on Multinomial Naive Bayes using t.f.-i.d.f. vectorized text. </p></div></div></div></div><footer><div class="container"><div class="seperator"></div><div class="row"><div class="col-md-6 col-xs-6 footer-link-column"><div class="footer-link-column-row"><h4>Community</h4><ul><li><a href="//docs.prediction.io/install/" target="blank">Download</a></li><li><a href="//docs.prediction.io/" target="blank">Docs</a></li><li><a href="//github.com/apache/incubator-predictionio" target="blank">GitHub</a></li><li><a href="mailto:[email protected]"; target="blank">Subscribe to User Mailing List</a></li ><li><a href="//stackoverflow.com/questions/tagged/predictionio" >target="blank">Stackoverflow</a></li></ul></div></div><div class="col-md-6 >col-xs-6 footer-link-column"><div >class="footer-link-column-row"><h4>Contribute</h4><ul><li><a >href="//predictionio.incubator.apache.org/community/contribute-code/" >target="blank">Contribute</a></li><li><a >href="//github.com/apache/incubator-predictionio" target="blank">Source >Code</a></li><li><a href="//issues.apache.org/jira/browse/PIO" >target="blank">Bug Tracker</a></li><li><a >href="mailto:[email protected]"; >target="blank">Subscribe to Development Mailing >List</a></li></ul></div></div></div></div><div id="footer-bottom"><div >class="container"><div class="row"><div class="col-md-12"><div >id="footer-logo-wrapper"><img alt="PredictionIO" >src="/images/logos/logo-white-d1e9c6e6.png"/></div><div >id="social-icons-wrapper"><a class="github-button" >href="https://github.com/apache/incubator-predictionio"; data-style="mega" data-count-href="/apache/incubator-predictionio/stargazers" data-count-api="/repos/apache/incubator-predictionio#stargazers_count" data-count-aria-label="# stargazers on GitHub" aria-label="Star apache/incubator-predictionio on GitHub">Star</a> <a class="github-button" href="https://github.com/apache/incubator-predictionio/fork"; data-icon="octicon-git-branch" data-style="mega" data-count-href="/apache/incubator-predictionio/network" data-count-api="/repos/apache/incubator-predictionio#forks_count" data-count-aria-label="# forks on GitHub" aria-label="Fork apache/incubator-predictionio on GitHub">Fork</a> <script id="github-bjs" async="" defer="" src="https://buttons.github.io/buttons.js";></script><a href="//www.facebook.com/predictionio" target="blank"><img alt="PredictionIO on Twitter" src="/images/icons/twitter-ea9dc152.png"/></a> <a href="//twitter.com/predictionio" target="blank"><img alt="PredictionIO on Facebook" src="/images/icons/facebook-5c57939c.png"/></a> </div></div></di v></div></div></footer></div><script>(function(w,d,t,u,n,s,e){w['SwiftypeObject']=n;w[n]=w[n]||function(){ (w[n].q=w[n].q||[]).push(arguments);};s=d.createElement(t); e=d.getElementsByTagName(t)[0];s.async=1;s.src=u;e.parentNode.insertBefore(s,e); })(window,document,'script','//s.swiftypecdn.com/install/v1/st.js','_st'); http://git-wip-us.apache.org/repos/asf/incubator-predictionio-site/blob/c2c7b62a/machinelearning/modelingworkflow/index.html.gz ---------------------------------------------------------------------- diff --git a/machinelearning/modelingworkflow/index.html.gz b/machinelearning/modelingworkflow/index.html.gz index fab9149..b32cd20 100644 Binary files a/machinelearning/modelingworkflow/index.html.gz and b/machinelearning/modelingworkflow/index.html.gz differ
