commons-math/xdocs/userguide/random.xml

361 lines
16 KiB
XML

<?xml version="1.0"?>
<!--
Licensed to the Apache Software Foundation (ASF) under one or more
contributor license agreements. See the NOTICE file distributed with
this work for additional information regarding copyright ownership.
The ASF licenses this file to You under the Apache License, Version 2.0
(the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
<?xml-stylesheet type="text/xsl" href="./xdoc.xsl"?>
<!-- $Revision$ $Date$ -->
<document url="random.html">
<properties>
<title>The Commons Math User Guide - Data Generation</title>
</properties>
<body>
<section name="2 Data Generation">
<subsection name="2.1 Overview" href="overview">
<p>
The Commons Math random package includes utilities for
<ul>
<li>generating random numbers</li>
<li>generating random strings</li>
<li>generating cryptographically secure sequences of random numbers or
strings</li>
<li>generating random samples and permuations</li>
<li>analyzing distributions of values in an input file and generating
values "like" the values in the file</li>
<li>generating data for grouped frequency distributions or
histograms</li>
</ul></p>
<p>
The source of random data used by the data generation utilities is
pluggable. By default, the JDK-supplied PseudoRandom Number Generator
(PRNG) is used, but alternative generators can be "plugged in" using an
adaptor framework, which provides a generic facility for replacing
<code>java.util.Random</code> with an alternative PRNG.
</p>
<p>
Sections 2.2-2.5 below show how to use the commons math API to generate
different kinds of random data. The examples all use the default
JDK-supplied PRNG. PRNG pluggability is covered in 2.6. The only
modification required to the examples to use alternative PRNGs is to
replace the argumentless constructor calls with invocations including
a <code>RandomGenerator</code> instance as a parameter.
</p>
</subsection>
<subsection name="2.2 Random numbers" href="deviates">
<p>
The <a href="../apidocs/org/apache/commons/math/random/RandomData.html">
org.apache.commons.math.RandomData</a> interface defines methods for
generating random sequences of numbers. The API contracts of these methods
use the following concepts:
<dl>
<dt>Random sequence of numbers from a probability distribution</dt>
<dd>There is no such thing as a single "random number." What can be
generated are <i>sequences</i> of numbers that appear to be random. When
using the built-in JDK function <code>Math.random(),</code> sequences of
values generated follow the
<a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda3662.htm">
Uniform Distribution</a>, which means that the values are evenly spread
over the interval between 0 and 1, with no sub-interval having a greater
probability of containing generated values than any other interval of the
same length. The mathematical concept of a
<a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda36.htm">
probability distribution</a> basically amounts to asserting that different
ranges in the set of possible values of a random variable have
different probabilities of containing the value. Commons Math supports
generating random sequences from the following probability distributions.
The javadoc for the <code>nextXxx</code> methods in
<code>RandomDataImpl</code> describes the algorithms used to generate
random deviates from each of these distributions.
<ul>
<li><a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda3662.htm">
uniform distribution</a></li>
<li><a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda3667.htm">
exponential distribution</a></li>
<li><a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda366j.htm">
poisson distribution</a></li>
<li><a href="http://www.itl.nist.gov/div898/handbook/eda/section3/eda3661.htm">
Gaussian distribution</a></li>
</ul>
</dd>
<dt>Cryptographically secure random sequences</dt>
<dd>It is possible for a sequence of numbers to appear random, but
nonetheless to be predictable based on the algorithm used to generate the
sequence. If in addition to randomness, strong unpredictability is
required, it is best to use a
<a href="http://www.wikipedia.org/wiki/Cryptographically_secure_pseudo-random_number_generator">
secure random number generator</a> to generate values (or strings). The
nextSecureXxx methods in the <code>RandomDataImpl</code> implementation of
the <code>RandomData</code> interface use the JDK <code>SecureRandom</code>
PRNG to generate cryptographically secure sequences. The
<code>setSecureAlgorithm</code> method allows you to change the underlying
PRNG. These methods are <strong>much slower</strong> than the corresponding
"non-secure" versions, so they should only be used when cryptographic
security is required.</dd>
<dt>Seeding pseudo-random number generators</dt>
<dd>By default, the implementation provided in <code>RandomDataImpl</code>
uses the JDK-provided PRNG. Like most other PRNGs, the JDK generator
generates sequences of random numbers based on an initial "seed value".
For the non-secure methods, starting with the same seed always produces the
same sequence of values. Secure sequences started with the same seeds will
diverge. When a new <code>RandomDataImpl</code> is created, the underlying
random number generators are <strong>not</strong> intialized. The first
call to a data generation method, or to a <code>reSeed()</code> method
initializes the appropriate generator. If you do not explicitly seed the
generator, it is by default seeded with the current time in milliseconds.
Therefore, to generate sequences of random data values, you should always
instantiate <strong>one</strong> <code>RandomDataImpl</code> and use it
repeatedly instead of creating new instances for subsequent values in the
sequence. For example, the following will generate a random sequence of 50
long integers between 1 and 1,000,000, using the current time in
milliseconds as the seed for the JDK PRNG:
<source>
RandomData randomData = new RandomDataImpl();
for (int i = 0; i &lt; 1000; i++) {
value = randomData.nextLong(1, 1000000);
}
</source>
The following will not in general produce a good random sequence, since the
PRNG is reseeded each time through the loop with the current time in
milliseconds:
<source>
for (int i = 0; i &lt; 1000; i++) {
RandomDataImpl randomData = new RandomDataImpl();
value = randomData.nextLong(1, 1000000);
}
</source>
The following will produce the same random sequence each time it is
executed:
<source>
RandomData randomData = new RandomDataImpl();
randomData.reSeed(1000);
for (int i = 0; i = 1000; i++) {
value = randomData.nextLong(1, 1000000);
}
</source>
The following will produce a different random sequence each time it is
executed.
<source>
RandomData randomData = new RandomDataImpl();
randomData.reSeedSecure(1000);
for (int i = 0; i &lt; 1000; i++) {
value = randomData.nextSecureLong(1, 1000000);
}
</source>
</dd></dl>
</p>
</subsection>
<subsection name="2.3 Random Strings" href="strings">
<p>
The methods <code>nextHexString</code> and <code>nextSecureHexString</code>
can be used to generate random strings of hexadecimal characters. Both
of these methods produce sequences of strings with good dispersion
properties. The difference between the two methods is that the second is
cryptographically secure. Specifically, the implementation of
<code>nextHexString(n)</code> in <code>RandomDataImpl</code> uses the
following simple algorithm to generate a string of <code>n</code> hex digits:
<ol>
<li>n/2+1 binary bytes are generated using the underlying Random</li>
<li>Each binary byte is translated into 2 hex digits</li></ol>
The <code>RandomDataImpl</code> implementation of the "secure" version,
<code>nextSecureHexString</code> generates hex characters in 40-byte
"chunks" using a 3-step process:
<ol>
<li>20 random bytes are generated using the underlying
<code>SecureRandom.</code></li>
<li>SHA-1 hash is applied to yield a 20-byte binary digest.</li>
<li>Each byte of the binary digest is converted to 2 hex digits</li></ol>
Similarly to the secure random number generation methods,
<code>nextSecureHexString</code> is <strong>much slower</strong> than
the non-secure version. It should be used only for applications such as
generating unique session or transaction ids where predictability of
subsequent ids based on observation of previous values is a security
concern. If all that is needed is an even distribution of hex characters
in the generated strings, the non-secure method should be used.
</p>
</subsection>
<subsection name="2.4 Random permutations, combinations, sampling"
href="combinatorics">
<p>
To select a random sample of objects in a collection, you can use the
<code>nextSample</code> method in the <code>RandomData</code> interface.
Specifically, if <code>c</code> is a collection containing at least
<code>k</code> objects, and <code>ranomData</code> is a
<code>RandomData</code> instance <code>randomData.nextSample(c, k)</code>
will return an <code>object[]</code> array of length <code>k</code>
consisting of elements randomly selected from the collection. If
<code>c</code> contains duplicate references, there may be duplicate
references in the returned array; otherwise returned elements will be
unique -- i.e., the sampling is without replacement among the object
references in the collection. </p>
<p>
If <code>randomData</code> is a <code>RandomData</code> instance, and
<code>n</code> and <code>k</code> are integers with
<code> k &lt;= n</code>, then
<code>randomData.nextPermutation(n, k)</code> returns an <code>int[]</code>
array of length <code>k</code> whose whose entries are selected randomly,
without repetition, from the integers <code>0</code> through
<code>n-1</code> (inclusive), i.e.,
<code>randomData.nextPermutation(n, k)</code> returns a random
permutation of <code>n</code> taken <code>k</code> at a time.
</p>
</subsection>
<subsection name="2.5 Generating data 'like' an input file" href="empirical">
<p>
Using the <code>ValueServer</code> class, you can generate data based on
the values in an input file in one of two ways:
<dl>
<dt>Replay Mode</dt>
<dd> The following code will read data from <code>url</code>
(a <code>java.net.URL</code> instance), cycling through the values in the
file in sequence, reopening and starting at the beginning again when all
values have been read.
<source>
ValueServer vs = new ValueServer();
vs.setValuesFileURL(url);
vs.setMode(ValueServer.REPLAY_MODE);
vs.resetReplayFile();
double value = vs.getNext();
// ...Generate and use more values...
vs.closeReplayFile();
</source>
The values in the file are not stored in memory, so it does not matter
how large the file is, but you do need to explicitly close the file
as above. The expected file format is \n -delimited (i.e. one per line)
strings representing valid floating point numbers.
</dd>
<dt>Digest Mode</dt>
<dd>When used in Digest Mode, the ValueServer reads the entire input file
and estimates a probability density function based on data from the file.
The estimation method is essentially the
<a href="http://nedwww.ipac.caltech.edu/level5/March02/Silverman/Silver2_6.html">
Variable Kernel Method</a> with Gaussian smoothing. Once the density
has been estimated, <code>getNext()</code> returns random values whose
probability distribution matches the empirical distribution -- i.e., if
you generate a large number of such values, their distribution should
"look like" the distribution of the values in the input file. The values
are not stored in memory in this case either, so there is no limit to the
size of the input file. Here is an example:
<source>
ValueServer vs = new ValueServer();
vs.setValuesFileURL(url);
vs.setMode(ValueServer.DIGEST_MODE);
vs.computeDistribution(500); //Read file and estimate distribution using 500 bins
double value = vs.getNext();
// ...Generate and use more values...
</source>
See the javadoc for <code>ValueServer</code> and
<code>EmpiricalDistribution</code> for more details. Note that
<code>computeDistribution()</code> opens and closes the input file
by itself.
</dd>
</dl>
</p>
</subsection>
<subsection name="2.6 PRNG Pluggability" href="pluggability">
<p>
To enable alternative PRNGs to be "plugged in" to the commons-math data
generation utilities and to provide a generic means to replace
<code>java.util.Random</code> in applications, a random generator
adaptor framework has been added to commons-math. The
<a href="../apidocs/org/apache/commons/math/random/RandomGenerator.html">
org.apache.commons.math.RandomGenerator</a> interface abstracts the public
interface of <code>java.util.Random</code> and any implementation of this
interface can be used as the source of random data for the commons-math
data generation classes. An abstract base class,
<a href="../apidocs/org/apache/commons/math/random/AbstractRandomGenerator.html">
org.apache.commons.math.AbstractRandomGenerator</a> is provided to make
implementation easier. This class provides default implementations of
"derived" data generation methods based on the primitive,
<code>nextDouble().</code> To support generic replacement of
<code>java.util.Random</code>, the
<a href="../apidocs/org/apache/commons/math/random/RandomAdaptor.html">
org.apache.commons.math.RandomAdaptor</a> class is provided, which
extends <code>java.util.Random</code> and wraps and delegates calls to
a <code>RandomGenerator</code> instance.
</p>
<p>
Examples:
<dl>
<dt>Create a RandomGenerator based on RngPack's Mersenne Twister</dt>
<dd>To create a RandomGenerator using the RngPack Mersenne Twister PRNG
as the source of randomness, extend <code>AbstractRandomGenerator</code>
overriding the derived methods that the RngPack implementation provides:
<source>
import edu.cornell.lassp.houle.RngPack.RanMT;
/**
* AbstractRandomGenerator based on RngPack RanMT generator.
*/
public class RngPackGenerator extends AbstractRandomGenerator {
private RanMT random = new RanMT();
public void setSeed(long seed) {
random = new RanMT(seed);
}
public double nextDouble() {
return random.raw();
}
public double nextGaussian() {
return random.gaussian();
}
public int nextInt(int n) {
return random.choose(n);
}
public boolean nextBoolean() {
return random.coin();
}
}
</source>
</dd>
<dt>Use the Mersenne Twister RandomGenerator in place of
<code>java.util.Random</code> in <code>RandomData</code></dt>
<dd>
<source>
RandomData randomData = new RandomDataImpl(new RngPackGenerator());
</source>
</dd>
<dt>Create an adaptor instance based on the Mersenne Twister generator
that can be used in place of a <code>Random</code></dt>
<dd>
<source>
RandomGenerator generator = new RngPackGenerator();
Random random = RandomAdaptor.createAdaptor(generator);
// random can now be used in place of a Random instance, data generation
// calls will be delegated to the wrapped Mersenne Twister
</source>
</dd>
</dl>
</p>
</subsection>
</section>
</body>
</document>