1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
|
---
layout: global
type: "page singular"
title: MLlib
description: MLlib is Apache Spark's scalable machine learning library, with APIs in Java, Scala and Python.
subproject: MLlib
---
<div class="jumbotron">
<b>MLlib</b> is Apache Spark's scalable machine learning library.
</div>
<div class="row row-padded">
<div class="col-md-7 col-sm-7">
<h2>Ease of Use</h2>
<p class="lead">
Usable in Java, Scala, Python, and SparkR.
</p>
<p>
MLlib fits into <a href="{{site.url}}">Spark</a>'s
APIs and interoperates with <a href="http://www.numpy.org">NumPy</a> in Python (starting in Spark 0.9).
You can use any Hadoop data source (e.g. HDFS, HBase, or local files), making it
easy to plug into Hadoop workflows.
</p>
</div>
<div class="col-md-5 col-sm-5 col-padded-top col-center">
<div style="margin-top: 15px; text-align: left; display: inline-block;">
<div class="code">
points = spark.textFile(<span class="string">"hdfs://..."</span>)<br/>
.<span class="sparkop">map</span>(<span class="closure">parsePoint</span>)<br/>
<br/>
model = KMeans.<span class="sparkop">train</span>(points, k=10)
</div>
<div class="caption">Calling MLlib in Python</div>
</div>
</div>
</div>
<div class="row row-padded">
<div class="col-md-7 col-sm-7">
<h2>Performance</h2>
<p class="lead">
High-quality algorithms, 100x faster than MapReduce.
</p>
<p>
Spark excels at iterative computation, enabling MLlib to run fast.
At the same time, we care about algorithmic performance:
MLlib contains high-quality algorithms that leverage iteration, and
can yield better results than the one-pass approximations sometimes used on MapReduce.
</p>
</div>
<div class="col-md-5 col-sm-5 col-padded-top col-center">
<div style="width: 100%; max-width: 272px; display: inline-block; text-align: center;">
<img src="{{site.url}}images/logistic-regression.png" style="width: 100%; max-width: 250px;">
<div class="caption" style="min-width: 272px;">Logistic regression in Hadoop and Spark</div>
</div>
</div>
</div>
<div class="row row-padded" style="margin-bottom: 15px;">
<div class="col-md-7 col-sm-7">
<h2>Easy to Deploy</h2>
<p class="lead">
Runs on existing Hadoop clusters and data.
</p>
<p>
If you have a Hadoop 2 cluster, you can run Spark and MLlib without any pre-installation.
Otherwise, Spark is easy to run <a href="{{site.url}}docs/latest/spark-standalone.html">standalone</a>
or on <a href="{{site.url}}docs/latest/ec2-scripts.html">EC2</a> or <a href="http://mesos.apache.org">Mesos</a>.
You can read from <a href="http://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-hdfs/HdfsUserGuide.html">HDFS</a>, <a href="http://hbase.apache.org">HBase</a>, or any Hadoop data source.
</p>
</div>
<div class="col-md-5 col-sm-5 col-padded-top col-center">
<img src="{{site.url}}images/hadoop.jpg" style="width: 100%; max-width: 280px;">
</div>
</div>
{% extra %}
<div class="row">
<div class="col-md-4 col-padded">
<h3>Algorithms</h3>
<p>
MLlib contains the following algorithms and utilities:
</p>
<ul class="list-narrow">
<li>logistic regression and linear support vector machine (SVM)</li>
<li>classification and regression tree</li>
<li>random forest and gradient-boosted trees</li>
<li>recommendation via alternating least squares (ALS)</li>
<li>clustering via k-means, Gaussian mixtures (GMM), and power iteration clustering</li>
<li>topic modeling via latent Dirichlet allocation (LDA)</li>
<li>singular value decomposition (SVD) and QR decomposition</li>
<li>principal component analysis (PCA)</li>
<li>linear regression with L<sub>1</sub>, L<sub>2</sub>, and elastic-net regularization</li>
<li>isotonic regression</li>
<li>multinomial/binomial naive Bayes</li>
<li>frequent itemset mining via FP-growth and association rules</li>
<li>sequential pattern mining via PrefixSpan</li>
<li>summary statistics and hypothesis testing</li>
<li>feature transformations</li>
<li>model evaluation and hyper-parameter tuning</li>
</ul>
<p>Refer to the <a href="{{site.url}}docs/latest/mllib-guide.html">MLlib guide</a> for usage examples.</p>
</div>
<div class="col-md-4 col-padded">
<h3>Community</h3>
<p>
MLlib is developed as part of the Apache Spark project. It thus gets
tested and updated with each Spark release.
</p>
<p>
If you have questions about the library, ask on the
<a href="{{site.url}}community.html#mailing-lists">Spark mailing lists</a>.
</p>
<p>
MLlib is still a young project and welcomes contributions. If you'd like to submit an algorithm to MLlib,
read <a href="https://cwiki.apache.org/confluence/display/SPARK/Contributing+to+Spark">how to
contribute to Spark</a> and send us a patch!
</p>
</div>
<div class="col-md-4 col-padded">
<h3>Getting Started</h3>
<p>
To get started with MLlib:
</p>
<ul class="list-narrow">
<li><a href="{{site.url}}downloads.html">Download Spark</a>. MLlib is included as a module.</li>
<li>Read the <a href="{{site.url}}docs/latest/mllib-guide.html">MLlib guide</a>, which includes
various usage examples.</li>
<li>Learn how to <a href="{{site.url}}docs/latest/#launching-on-a-cluster">deploy</a> Spark on a cluster
if you'd like to run in distributed mode. You can also run locally on a multicore machine
without any setup.
</li>
</ul>
</div>
</div>
<div class="row">
<div class="col-sm-12 col-center">
<a href="{{site.url}}downloads.html" class="btn btn-success btn-lg btn-multiline">
Download Spark<br/><span class="small">Includes MLlib</span>
</a>
</div>
</div>
{% endextra %}
|