blob: 9c15ecc5f0c9d958639ce4b4dae82cbca47edbb7 [file] [log] [blame]
<!DOCTYPE html>
<html lang="en">
<head>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-61232409-1"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-61232409-1');
</script>
<meta charset="UTF-8">
<meta name="ignite-version" content="2.9.0" />
<title>Gradient Boosting | Ignite Documentation</title>
<link rel="canonical" href="/docs/latest/machine-learning/ensemble-methods/gradient-boosting" />
<link rel="stylesheet" href="/assets/css/styles.css?1600975887">
<link rel="stylesheet" href="/assets/css/asciidoc-pygments.css">
<link rel="shortcut icon" href="/favicon.ico">
<meta name='viewport' content='width=device-width, height=device-height, initial-scale=1.0, minimum-scale=1.0'>
<script type="text/javascript" src="/assets/js/anchor.min.js?1600975887"></script>
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/docsearch.js@2/dist/cdn/docsearch.min.css" />
</head>
<body>
<header>
<!--#include virtual="/includes/promotion_banner.html" -->
<div class="container">
<button type='button' class='menu' title='Docs menu'>
<img src="/assets/images/menu-icon.svg"/>
</button>
<div class='home'>
<a href="/" class='home' title='Apache Ignite home'>
<img src="/assets/images/apache_ignite_logo.svg" alt="Apache Ignite logo" width="103" >
</a>
</div>
<nav>
</nav>
<select id="version-selector">
<option value="2.9.0">2.9.0</option>
</select>
<a href="https://github.com/apache/ignite" title='GitHub' class='github' target="_blank">
<img src="/assets/images/github-gray.svg" alt="GitHub logo">
</a>
<form class='search'>
<button class="search-close" type='button'><img src='/assets/images/cancel.svg'></button>
<input type="search" placeholder="Search…" id="search-input">
</form>
<button type='button' class='search-toggle'><img src='/assets/images/search.svg'></button>
<button type='button' class='top-nav-toggle'></button>
</div>
</header>
<link rel="stylesheet" href="/assets/css/docs.css">
<section class='page-docs'>
<nav class='left-nav' data-swiftype-index='false'>
<li>
<a href="/docs/latest/preface" class='' >Preface</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Quick Start Guides<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/quick-start/java"
class=''
>Java</a>
</li>
<li>
<a href="/docs/latest/quick-start/dotnet"
class=''
>.NET/C#</a>
</li>
<li>
<a href="/docs/latest/quick-start/cpp"
class=''
>C++</a>
</li>
<li>
<a href="/docs/latest/quick-start/python"
class=''
>Python</a>
</li>
<li>
<a href="/docs/latest/quick-start/nodejs"
class=''
>Node.JS</a>
</li>
<li>
<a href="/docs/latest/quick-start/sql"
class=''
>SQL</a>
</li>
<li>
<a href="/docs/latest/quick-start/php"
class=''
>PHP</a>
</li>
<li>
<a href="/docs/latest/quick-start/restapi"
class=''
>REST API</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="/installation" class='group-toggle collapsed '>Installation<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/installation/installing-using-zip"
class=''
>Installing Using ZIP Archive</a>
</li>
<li>
<a href="/docs/latest/installation/installing-using-docker"
class=''
>Installing Using Docker</a>
</li>
<li>
<a href="/docs/latest/installation/deb-rpm"
class=''
>Installing DEB or RPM package</a>
</li>
<li>
<button
type='button'
class='collapsed '>Kubernetes<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//installation/kubernetes/amazon-eks-deployment" class=''>Amazon EKS</a></li>
<li><a href="/docs/latest//installation/kubernetes/azure-deployment" class=''>Azure Kubernetes Service</a></li>
<li><a href="/docs/latest//installation/kubernetes/gke-deployment" class=''>Google Kubernetes Engine</a></li>
</nav>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Setting Up<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/setup"
class=''
>Setting Up Ignite for Java</a>
</li>
<li>
<a href="/docs/latest/setup-dotnet"
class=''
>Setting Up Ignite for .NET/C#</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/understanding-configuration" class='' >Understanding Configuration</a>
</li>
<li>
<a href="/docs/latest/logging" class='' >Configuring Logging</a>
</li>
<li>
<a href="/docs/latest/starting-nodes" class='' >Starting and Stopping Nodes</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Clustering<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/clustering/clustering"
class=''
>Overview</a>
</li>
<li>
<a href="/docs/latest/clustering/tcp-ip-discovery"
class=''
>TCP/IP Discovery</a>
</li>
<li>
<a href="/docs/latest/clustering/zookeeper-discovery"
class=''
>ZooKeeper Discovery</a>
</li>
<li>
<a href="/docs/latest/clustering/discovery-in-the-cloud"
class=''
>Discovery in the Cloud</a>
</li>
<li>
<a href="/docs/latest/clustering/network-configuration"
class=''
>Network Configuration</a>
</li>
<li>
<a href="/docs/latest/clustering/connect-client-nodes"
class=''
>Connecting Client Nodes</a>
</li>
<li>
<a href="/docs/latest/clustering/running-client-nodes-behind-nat"
class=''
>Running Client Nodes Behind NAT</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Data Modeling<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/data-modeling/data-modeling"
class=''
>Introduction</a>
</li>
<li>
<a href="/docs/latest/data-modeling/data-partitioning"
class=''
>Data Partitioning</a>
</li>
<li>
<a href="/docs/latest/data-modeling/affinity-collocation"
class=''
>Affinity Colocation</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Configuring Memory<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/memory-architecture"
class=''
>Memory Architecture</a>
</li>
<li>
<a href="/docs/latest/memory-configuration/data-regions"
class=''
>Configuring Data Regions</a>
</li>
<li>
<a href="/docs/latest/memory-configuration/eviction-policies"
class=''
>Eviction Policies</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Configuring Caches<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/configuring-caches/configuration-overview"
class=''
>Cache Configuration</a>
</li>
<li>
<a href="/docs/latest/configuring-caches/configuring-backups"
class=''
>Configuring Partition Backups</a>
</li>
<li>
<a href="/docs/latest/configuring-caches/atomicity-modes"
class=''
>Atomicity Modes</a>
</li>
<li>
<a href="/docs/latest/configuring-caches/expiry-policies"
class=''
>Expiry Policy</a>
</li>
<li>
<a href="/docs/latest/configuring-caches/on-heap-caching"
class=''
>On-Heap Caching</a>
</li>
<li>
<a href="/docs/latest/configuring-caches/cache-groups"
class=''
>Cache Groups</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Persistence<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/persistence/native-persistence"
class=''
>Ignite Persistence</a>
</li>
<li>
<a href="/docs/latest/persistence/external-storage"
class=''
>External Storage</a>
</li>
<li>
<a href="/docs/latest/persistence/swap"
class=''
>Swapping</a>
</li>
<li>
<a href="/docs/latest/persistence/custom-cache-store"
class=''
>Implementing Custom Cache Store</a>
</li>
<li>
<a href="/docs/latest/persistence/snapshots"
class=''
>Cluster Snapshots</a>
</li>
<li>
<a href="/docs/latest/persistence/disk-compression"
class=''
>Disk Compression</a>
</li>
<li>
<a href="/docs/latest/persistence/persistence-tuning"
class=''
>Tuning Persistence</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/baseline-topology" class='' >Baseline Topology</a>
</li>
<li>
<a href="/docs/latest/cluster-states" class='' >Cluster States</a>
</li>
<li>
<a href="/docs/latest/data-rebalancing" class='' >Data Rebalancing</a>
</li>
<li>
<a href="/docs/latest/partition-loss-policy" class='' >Partition Loss Policy</a>
</li>
<li>
<a href="/docs/latest/deploying-user-code" class='' >Deploying User Code</a>
</li>
<li>
<a href="/docs/latest/peer-class-loading" class='' >Peer Class Loading</a>
</li>
<li>
<a href="/docs/latest/data-streaming" class='' >Data Streaming</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Using Key-Value Cache API<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/key-value-api/basic-cache-operations"
class=''
>Basic Cache Operations</a>
</li>
<li>
<a href="/docs/latest/key-value-api/binary-objects"
class=''
>Working with Binary Objects</a>
</li>
<li>
<a href="/docs/latest/key-value-api/using-scan-queries"
class=''
>Using Scan Queries</a>
</li>
<li>
<a href="/docs/latest/read-repair"
class=''
>Read Repair</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/key-value-api/continuous-queries" class='' >Using Continuous Queries</a>
</li>
<li>
<a href="/docs/latest/key-value-api/transactions" class='' >Performing Transactions</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Working with SQL<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/SQL/sql-introduction"
class=''
>Introduction</a>
</li>
<li>
<a href="/docs/latest/SQL/schemas"
class=''
>Understanding Schemas</a>
</li>
<li>
<a href="/docs/latest/SQL/indexes"
class=''
>Defining Indexes</a>
</li>
<li>
<a href="/docs/latest/SQL/sql-api"
class=''
>Using SQL API</a>
</li>
<li>
<a href="/docs/latest/SQL/distributed-joins"
class=''
>Distributed Joins</a>
</li>
<li>
<a href="/docs/latest/SQL/sql-transactions"
class=''
>SQL Transactions</a>
</li>
<li>
<a href="/docs/latest/SQL/custom-sql-func"
class=''
>Custom SQL Functions</a>
</li>
<li>
<a href="/docs/latest/SQL/JDBC/jdbc-driver"
class=''
>JDBC Driver</a>
</li>
<li>
<a href="/docs/latest/SQL/JDBC/jdbc-client-driver"
class=''
>JDBC Client Driver</a>
</li>
<li>
<a href="/docs/latest/transactions/mvcc"
class=''
>Multiversion Concurrency Control</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Distributed Computing<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/distributed-computing/distributed-computing"
class=''
>Distributed Computing API</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/cluster-groups"
class=''
>Cluster Groups</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/executor-service"
class=''
>Executor Service</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/map-reduce"
class=''
>MapReduce API</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/load-balancing"
class=''
>Load Balancing</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/fault-tolerance"
class=''
>Fault Tolerance</a>
</li>
<li>
<a href="/docs/latest/distributed-computing/job-scheduling"
class=''
>Job Scheduling</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/collocated-computations" class='' >Colocating Computations with Data</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Working with Events<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/events/listening-to-events"
class=''
>Enabling and Listenting to Events</a>
</li>
<li>
<a href="/docs/latest/events/events"
class=''
>Events</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/near-cache" class='' >Near Caches</a>
</li>
<li>
<a href="/docs/latest/platform-cache" class='' >.NET Platform Cache</a>
</li>
<li>
<a href="/docs/latest/services/services" class='' >Services</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Distributed Data Structures<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/data-structures/queue-and-set"
class=''
>Queue and Set</a>
</li>
<li>
<a href="/docs/latest/data-structures/atomic-types"
class=''
>Atomic Types</a>
</li>
<li>
<a href="/docs/latest/data-structures/countdownlatch"
class=''
>CountDownLatch</a>
</li>
<li>
<a href="/docs/latest/data-structures/atomic-sequence"
class=''
>Atomic Sequence</a>
</li>
<li>
<a href="/docs/latest/data-structures/semaphore"
class=''
>Semaphore</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Machine Learning<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/machine-learning/machine-learning"
class=''
>Machine Learning</a>
</li>
<li>
<a href="/docs/latest/machine-learning/partition-based-dataset"
class=''
>Partition Based Dataset</a>
</li>
<li>
<a href="/docs/latest/machine-learning/updating-trained-models"
class=''
>Updating Trained Models</a>
</li>
<li>
<button
type='button'
class='collapsed '>Binary Classification<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/binary-classification/introduction" class=''>Introduction</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/linear-svm" class=''>Linear SVM (Support Vector Machine)</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/decision-trees" class=''>Decision Trees</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/multilayer-perceptron" class=''>Multilayer Perceptron</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/logistic-regression" class=''>Logistic Regression</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/knn-classification" class=''>k-NN Classification</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/ann" class=''>ANN (Approximate Nearest Neighbor)</a></li>
<li><a href="/docs/latest//machine-learning/binary-classification/naive-bayes" class=''>Naive Bayes</a></li>
</nav>
</li>
<li>
<button
type='button'
class='collapsed '>Regression<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/regression/introduction" class=''>Introduction</a></li>
<li><a href="/docs/latest//machine-learning/regression/linear-regression" class=''>Linear Regression</a></li>
<li><a href="/docs/latest//machine-learning/regression/decision-trees-regression" class=''>Decision Trees Regression</a></li>
<li><a href="/docs/latest//machine-learning/regression/knn-regression" class=''>k-NN Regression</a></li>
</nav>
</li>
<li>
<button
type='button'
class='collapsed '>Clustering<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/clustering/introduction" class=''>Introduction</a></li>
<li><a href="/docs/latest//machine-learning/clustering/k-means-clustering" class=''>K-Means Clustering</a></li>
<li><a href="/docs/latest//machine-learning/clustering/gaussian-mixture" class=''>Gaussian mixture (GMM)</a></li>
</nav>
</li>
<li>
<a href="/docs/latest/machine-learning/preprocessing"
class=''
>Preprocessing</a>
</li>
<li>
<button
type='button'
class='collapsed '>Model Selection<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/model-selection/introduction" class=''>Introduction</a></li>
<li><a href="/docs/latest//machine-learning/model-selection/evaluator" class=''>Evaluator</a></li>
<li><a href="/docs/latest//machine-learning/model-selection/split-the-dataset-on-test-and-train-datasets" class=''>Split the dataset on test and train datasets</a></li>
<li><a href="/docs/latest//machine-learning/model-selection/hyper-parameter-tuning" class=''>Hyper-parameter tuning</a></li>
<li><a href="/docs/latest//machine-learning/model-selection/pipeline-api" class=''>Pipeline API</a></li>
</nav>
</li>
<li>
<a href="/docs/latest/machine-learning/multiclass-classification"
class=''
>Multiclass Classification</a>
</li>
<li>
<button
type='button'
class='collapsed '>Ensemble Methods<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/ensemble-methods/introduction" class=''></a></li>
<li><a href="/docs/latest//machine-learning/ensemble-methods/stacking" class=''>Stacking</a></li>
<li><a href="/docs/latest//machine-learning/ensemble-methods/baggin" class=''>Bagging</a></li>
<li><a href="/docs/latest//machine-learning/ensemble-methods/random-forest" class=''>Random Forest</a></li>
<li><a href="/docs/latest//machine-learning/ensemble-methods/gradient-boosting" class=''>Gradient Boosting</a></li>
</nav>
</li>
<li>
<a href="/docs/latest/machine-learning/recommendation-systems"
class=''
>Recommendation Systems</a>
</li>
<li>
<button
type='button'
class='collapsed '>Importing Model<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//machine-learning/importing-model/introduction" class=''>Introduction</a></li>
<li><a href="/docs/latest//machine-learning/importing-model/model-import-from-gxboost" class=''>Import Model from XGBoost</a></li>
<li><a href="/docs/latest//machine-learning/importing-model/model-import-from-apache-spark" class=''>Import Model from Apache Spark</a></li>
</nav>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Monitoring<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/monitoring-metrics/intro"
class=''
>Introduction</a>
</li>
<li>
<a href="/docs/latest/monitoring-metrics/cluster-id"
class=''
>Cluster ID and Tag</a>
</li>
<li>
<button
type='button'
class='collapsed '>Metrics<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//monitoring-metrics/configuring-metrics" class=''>Configuring Metrics</a></li>
<li><a href="/docs/latest//monitoring-metrics/metrics" class=''>JMX Metrics</a></li>
</nav>
</li>
<li>
<button
type='button'
class='collapsed '>New Metrics System<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//monitoring-metrics/new-metrics-system" class=''>Introduction</a></li>
<li><a href="/docs/latest//monitoring-metrics/new-metrics" class=''>Metrics</a></li>
</nav>
</li>
<li>
<a href="/docs/latest/monitoring-metrics/system-views"
class=''
>System Views</a>
</li>
<li>
<a href="/docs/latest/monitoring-metrics/tracing"
class=''
>Tracing</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="/security" class='group-toggle collapsed '>Security<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/security/authentication"
class=''
>Authentication</a>
</li>
<li>
<a href="/docs/latest/security/ssl-tls"
class=''
>SSL/TLS</a>
</li>
<li>
<button
type='button'
class='collapsed '>Transparent Data Encryption<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//security/tde" class=''>Introduction</a></li>
<li><a href="/docs/latest//security/master-key-rotation" class=''>Master key rotation</a></li>
</nav>
</li>
<li>
<a href="/docs/latest/security/sandbox"
class=''
>Sandbox</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Thin Clients<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/thin-clients/getting-started-with-thin-clients"
class=''
>Thin Clients Overview</a>
</li>
<li>
<a href="/docs/latest/thin-clients/java-thin-client"
class=''
>Java Thin Client</a>
</li>
<li>
<a href="/docs/latest/thin-clients/dotnet-thin-client"
class=''
>.NET Thin Client</a>
</li>
<li>
<a href="/docs/latest/thin-clients/cpp-thin-client"
class=''
>C++ Thin Client</a>
</li>
<li>
<a href="/docs/latest/thin-clients/python-thin-client"
class=''
>Python Thin Client</a>
</li>
<li>
<a href="/docs/latest/thin-clients/php-thin-client"
class=''
>PHP Thin Client</a>
</li>
<li>
<a href="/docs/latest/thin-clients/nodejs-thin-client"
class=''
>Node.js Thin Client</a>
</li>
<li>
<button
type='button'
class='collapsed '>Binary Client Protocol<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class="sub_pages collapsed">
<li><a href="/docs/latest//binary-client-protocol/binary-client-protocol" class=''>Binary Client Protocol</a></li>
<li><a href="/docs/latest//binary-client-protocol/data-format" class=''>Data Format</a></li>
<li><a href="/docs/latest//binary-client-protocol/key-value-queries" class=''>Key-Value Queries</a></li>
<li><a href="/docs/latest//binary-client-protocol/sql-and-scan-queries" class=''>SQL and Scan Queries</a></li>
<li><a href="/docs/latest//binary-client-protocol/binary-type-metadata" class=''>Binary Types Metadata</a></li>
<li><a href="/docs/latest//binary-client-protocol/cache-configuration" class=''>Cache Configuration</a></li>
</nav>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>ODBC Driver<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/SQL/ODBC/odbc-driver"
class=''
>ODBC Driver</a>
</li>
<li>
<a href="/docs/latest/SQL/ODBC/connection-string-dsn"
class=''
>Connection String and DSN</a>
</li>
<li>
<a href="/docs/latest/SQL/ODBC/querying-modifying-data"
class=''
>Querying and Modifying Data</a>
</li>
<li>
<a href="/docs/latest/SQL/ODBC/specification"
class=''
>Specification</a>
</li>
<li>
<a href="/docs/latest/SQL/ODBC/data-types"
class=''
>Data Types</a>
</li>
<li>
<a href="/docs/latest/SQL/ODBC/error-codes"
class=''
>Error Codes</a>
</li>
</nav>
</li>
<li>
<a href="/docs/latest/restapi" class='' >REST API</a>
</li>
<li>
<a href="/docs/latest/control-script" class='' >Control Script</a>
</li>
<li>
<a href="/docs/latest/plugins" class='' >Plugins</a>
</li>
<li>
<a href="/docs/latest/sqlline" class='' >SQLLine</a>
</li>
<li>
<button type='button' data-guide-url="" class='group-toggle collapsed '>Ignite for Spark<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/ignite-for-spark/overview"
class=''
>Overview</a>
</li>
<li>
<a href="/docs/latest/ignite-for-spark/ignitecontext-and-rdd"
class=''
>IgniteContext and IgniteRDD</a>
</li>
<li>
<a href="/docs/latest/ignite-for-spark/ignite-dataframe"
class=''
>Ignite DataFrame</a>
</li>
<li>
<a href="/docs/latest/ignite-for-spark/installation"
class=''
>Installation</a>
</li>
<li>
<a href="/docs/latest/ignite-for-spark/spark-shell"
class=''
>Test Ignite with Spark-shell</a>
</li>
<li>
<a href="/docs/latest/ignite-for-spark/troubleshooting"
class=''
>Troubleshooting</a>
</li>
</nav>
</li>
<li>
<button type='button' data-guide-url="/sql-reference/sql-reference-overview" class='group-toggle collapsed '>SQL Reference<img class="state-indicator" src="/assets/images/left-nav-arrow.svg"></button>
<nav class='nav-group collapsed'>
<li>
<a href="/docs/latest/sql-reference/sql-conformance"
class=''
>SQL Conformance</a>
</li>
<li>
<a href="/docs/latest/sql-reference/ddl"
class=''
>Data Definition Language (DDL)</a>
</li>
<li>
<a href="/docs/latest/sql-reference/dml"
class=''
>Data Manipulation Language (DML)</a>
</li>
<li>
<a href="/docs/latest/sql-reference/transactions"
class=''
>Transactions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/operational-commands"
class=''
>Operational Commands</a>
</li>
<li>
<a href="/docs/latest/sql-reference/aggregate-functions"
class=''
>Aggregate functions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/numeric-functions"
class=''
>Numeric Functions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/string-functions"
class=''
>String Functions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/date-time-functions"
class=''
>Data and Time Functions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/system-functions"
class=''
>System Functions</a>
</li>
<li>
<a href="/docs/latest/sql-reference/data-types"
class=''
>Data Types</a>
</li>
</nav>
</li>
</nav>
<div class="left-nav__overlay"></div>
<article data-swiftype-index='true'>
<a class='edit-link' href="https://github.com/apache/ignite/tree/IGNITE-7595/docs/_docs/machine-learning/ensemble-methods/gradient-boosting.adoc" target="_blank">Edit</a>
<h1>Gradient Boosting</h1>
<div id="preamble">
<div class="sectionbody">
<div class="paragraph">
<p>In machine learning, boosting is an ensemble meta-algorithm for primarily reducing bias, and also variance in supervised learning, and a family of machine learning algorithms that convert weak learners to strong ones.</p>
</div>
<div class="admonitionblock note">
<table>
<tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">
<h3 id="question-posed-by-kearns-and-valiant-1988-1989" class="discrete">Question posed by Kearns and Valiant (1988, 1989)</h3>
<div class="paragraph">
<p>"Can a set of weak learners create a single strong learner?"</p>
</div>
<div class="literalblock">
<div class="content">
<pre>A weak learner is defined to be a classifier that is only slightly correlated with the true classification (it can label examples better than random guessing). In contrast, a strong learner is a classifier that is arbitrarily well-correlated with the true classification.</pre>
</div>
</div>
</td>
</tr>
</table>
</div>
<div class="paragraph">
<p>Later, in 1990 it was demonstrated by Robert Schapire and led to the boosting technique development.</p>
</div>
<div class="paragraph">
<p>The boosing is presented in Ignite ML library as a Gradient Boosting (the most popular boosting implementation).</p>
</div>
</div>
</div>
<div class="sect1">
<h2 id="overview">Overview</h2>
<div class="sectionbody">
<div class="paragraph">
<p>Gradient boosting is a machine learning technique that produces a prediction model in the form of an <a href="https://en.wikipedia.org/wiki/Ensemble_learning">ensemble</a> of weak prediction models. A gradient boosting algorithm tries to solve the minimization error problem on learning samples in a functional space where each function is a model. Each model in this composition tries to predict a gradient of error for points in a feature space and these predictions will be summed with some weight to model an answer. This algorithm may be used for regression and classification problems. For more information please see <a href="https://en.wikipedia.org/wiki/Gradient_boosting">Wikipedia</a>.</p>
</div>
<div class="paragraph">
<p>In Ignite ML there is an implementation of a general GDB algorithm and GDB-on-trees algorithm. General GDB (GDBRegressionTrainer and GDBBinaryClassifierTrainer) allows any trainer for training each model in composition. GDB on trees uses some optimizations specific for trees, such as indexes, for avoiding sorting during the decision tree build phase.</p>
</div>
</div>
</div>
<div class="sect1">
<h2 id="model">Model</h2>
<div class="sectionbody">
<div class="paragraph">
<p>Apache Ignite ML purposes all implementations of the GDB algorithm to use GDBModel, wrapping ModelsComposition for representing the composition of a few models. ModelsComposition implements a common Model interface and can be used as follows:</p>
</div>
<div class="listingblock">
<div class="content">
<pre class="rouge highlight"><code data-lang="java"><span class="nc">GDBModel</span> <span class="n">model</span> <span class="o">=</span> <span class="o">...;</span>
<span class="kt">double</span> <span class="n">prediction</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="na">predict</span><span class="o">(</span><span class="n">observation</span><span class="o">);</span></code></pre>
</div>
</div>
<div class="paragraph">
<p>GDBModel uses WeightedPredictionsAggregator as the model answer reducer. This aggregator computes an answer of a meta-model, since “result = bias + p1*w1 + p2*w2 + &#8230;&#8203;” where</p>
</div>
<div class="ulist">
<ul>
<li>
<p><code>pi</code> - answer of i-th model.</p>
</li>
<li>
<p><code>wi</code> - weight of model in composition.</p>
</li>
</ul>
</div>
<div class="paragraph">
<p>GDB uses the mean value of labels for the bias-parameter in the aggregator.</p>
</div>
</div>
</div>
<div class="sect1">
<h2 id="trainer">Trainer</h2>
<div class="sectionbody">
<div class="paragraph">
<p>Training of GDB is represented by <code>GDBRegressionTrainer</code>, <code>GDBBinaryClassificationTrainer</code> and <code>GDBRegressionOnTreesTrainer</code>, <code>GDBBinaryClassificationOnTreesTrainer</code> for general GDB and GDB on trees respectively. All trainers have the following parameters:</p>
</div>
<div class="ulist">
<ul>
<li>
<p><code>gradStepSize</code> - sets the constant weight of each model in composition; in future versions of Ignite ML this parameter may be computed dynamically.</p>
</li>
<li>
<p><code>cntOfIterations</code> - sets the maximum of models in the composition after training.</p>
</li>
<li>
<p><code>checkConvergenceFactory</code> - sets factory for construction of convergence checker used for preventing overfitting and learning of many useless models while training.</p>
</li>
</ul>
</div>
<div class="paragraph">
<p>For classifier trainers there is addition parameter:</p>
</div>
<div class="ulist">
<ul>
<li>
<p><code>loss</code> - sets loss computer on some learning example from a training dataset.</p>
</li>
</ul>
</div>
<div class="paragraph">
<p>There are several factories for convergence checkers:</p>
</div>
<div class="ulist">
<ul>
<li>
<p><code>ConvergenceCheckerStubFactory</code> creates a checker that always returns false for a convergence check. So in this case, model composition size will have cntOfIterations models.</p>
</li>
<li>
<p><code>MeanAbsValueConvergenceCheckerFactory</code> creates a checker that compute a mean value of the absolute gradient values on each example from a dataset and returns true if this it is less than the used-defined threshold.</p>
</li>
<li>
<p><code>MedianOfMedianConvergenceCheckerFactory</code> creates a checker that computes the median of median absolute gradient values on each data partition. This method is less sensitive for anomalies in the learning dataset, but GDB may converge longer.</p>
</li>
</ul>
</div>
<div class="paragraph">
<p>Example of training:</p>
</div>
<div class="listingblock">
<div class="content">
<pre class="rouge highlight"><code data-lang="java"><span class="c1">// Set up trainer</span>
<span class="nc">GDBTrainer</span> <span class="n">trainer</span> <span class="o">=</span> <span class="k">new</span> <span class="nc">GDBBinaryClassifierOnTreesTrainer</span><span class="o">(</span>
<span class="n">learningRate</span><span class="o">,</span> <span class="n">countOfIterations</span><span class="o">,</span> <span class="k">new</span> <span class="nc">LogLoss</span><span class="o">()</span>
<span class="o">).</span><span class="na">withCheckConvergenceStgyFactory</span><span class="o">(</span><span class="k">new</span> <span class="nc">MedianOfMedianConvergenceCheckFactory</span><span class="o">(</span><span class="n">precision</span><span class="o">));</span>
<span class="c1">// Build the model</span>
<span class="nc">GDBModel</span> <span class="n">mdl</span> <span class="o">=</span> <span class="n">trainer</span><span class="o">.</span><span class="na">fit</span><span class="o">(</span>
<span class="n">ignite</span><span class="o">,</span>
<span class="n">dataCache</span><span class="o">,</span>
<span class="n">vectorizer</span>
<span class="o">);</span></code></pre>
</div>
</div>
</div>
</div>
<div class="sect1">
<h2 id="example">Example</h2>
<div class="sectionbody">
<div class="paragraph">
<p>To see how GDB Classifier can be used in practice, try this <a href="https://github.com/apache/ignite/blob/master/examples/src/main/java/org/apache/ignite/examples/ml/tree/boosting/GDBOnTreesClassificationTrainerExample.java">example</a> that is available on GitHub and delivered with every Apache Ignite distribution.</p>
</div>
</div>
</div>
<div class="copyright">
© 2020 The Apache Software Foundation.<br/>
Apache, Apache Ignite, the Apache feather and the Apache Ignite logo are either registered trademarks or trademarks of The Apache Software Foundation.
</div>
</article>
<nav class="right-nav" data-swiftype-index='false'>
<ul class="sectlevel1">
<li><a href="#overview">Overview</a></li>
<li><a href="#model">Model</a></li>
<li><a href="#trainer">Trainer</a></li>
<li><a href="#example">Example</a></li>
</ul>
<footer>
</footer>
</nav>
</section>
<script type='module' src='/assets/js/code-copy-to-clipboard.js' async crossorigin></script>
<script>
// inits deep anchors -- needs to be done here because of https://www.bryanbraun.com/anchorjs/#dont-run-it-too-late
anchors.add('.page-docs h1, .page-docs h2, .page-docs h3:not(.discrete), .page-docs h4, .page-docs h5');
anchors.options = {
placement: 'right',
visible: 'always'
};
</script>
<!-- load google fonts async -->
<script type="text/javascript">
WebFontConfig = {
google: { families: [ 'Open+Sans:300,400,600,700&display=swap' ] }
};
(function() {
var wf = document.createElement('script');
wf.src = 'https://ajax.googleapis.com/ajax/libs/webfont/1/webfont.js';
wf.type = 'text/javascript';
wf.async = 'true';
var s = document.getElementsByTagName('script')[0];
s.parentNode.insertBefore(wf, s);
})(); </script>
<script src="https://cdn.jsdelivr.net/npm/docsearch.js@2/dist/cdn/docsearch.min.js"></script>
<script>
docsearch({
// Your apiKey and indexName will be given to you once
// we create your config
apiKey: '3eee686c0ebe39eff3baeb18c56fa5f8',
indexName: 'apache_ignite',
// Replace inputSelector with a CSS selector
// matching your search input
inputSelector: '#search-input',
// algoliaOptions: { 'facetFilters': ["version:$VERSION"] },
// Set debug to true to inspect the dropdown
debug: false,
});
</script>
<script type='module' src='/assets/js/index.js?1600975887' async crossorigin></script>
<script type='module' src='/assets/js/versioning.js?1600975887' async crossorigin></script>
</body>
</html>