Home

Awesome

<img src="http://numenta.org/87b23beb8a4b7dea7d88099bfb28d182.svg" alt="NuPIC Logo" width=100/>

htm.java

<br>

htm.java awesomeness AGI Probability Coolness Factor Build Status Coverage Status Maven Central docs-badge Gitter OpenHub

<br>

Official Java™ version of...

Hierarchical Temporal Memory (HTM)

Community-supported & ported from the Numenta Platform for Intelligent Computing (NuPIC) python project.

NOTE: Minimum JavaSE version is 8

<br>

For Demos & Examples: see the HTM.java-examples repository


<br>

Recent News Items...

News Archives...

View the Change Log (Updated! 2017-04-05)


For a more detailed discussion of <b>htm.java</b> see: <BR>

See the Test Coverage Reports - For more information on where you can contribute! Extend the tests and get your name in bright lights!

For answers to more detailed questions, post to the HTM Forum (can be used via email too), or chat with us on Gitter.

Gitter

For more detailed discussions regarding HTM.java specifically, come chat with us here:

Gitter

See the blog: Join the Cogmission

<a name="callToArms"></a>

Call to Arms: HTM.java needs you!


Versioning

(Tracked according to core algorithms)

Core AlgorithmNuPIC DateHTM.Java DateLatest NuPIC SHALatest HTM.Java SHAStatus
SpatialPooler2016-12-112016-10-07commitcommit*Behind NuPIC Merge #3411
TemporalMemory2017-06-022016-10-13commitcommit*Behind NuPIC Merge #3654

* May be one of: "Sync'd" or "Behind". "Behind" expresses a temporary lapse in synchronization while devs are implementing new changes.

<sub><sup>NOTE: "Behind" status does not imply any lack of operational ability. The master branch of HTM.Java will always be fully operational. <br> Any fully critical feature addition to NuPIC will always be matched (sync'd) ASAP, however due to ongoing updates, the algortithms within NuPIC may reach ahead for a short period of time while the HTM Community is busy porting the difference(s). We are committed to keeping HTM.Java up to date with NuPIC at all times.</sup></sub>


Project Goals

The primary goal of this library development is to provide a Java version of NuPIC that has a 1-to-1 correspondence to all systems, functionality and tests provided by Numenta's open source implementation; while observing the tenets, standards and conventions of Java language best practices and development.

By working closely with Numenta and receiving their enthusiastic support and guidance, it is intended that this library be maintained as a viable Java language alternative to Numenta's C++ and Python offerings. However it must be understood that "official" support is (for the time being) currently limited to community resources such as the maintainers of this library and Numenta Forums / Message Lists and IRC:


How Do I Get The Code?

A. Developers who wish to make contributions are required to Fork the htm.java repo and then clone from their personal "Fork" of htm.java...

your_git_directory% git clone https://github.com/<your_github_username>/htm.java.git

B. Anybody who just wants to "play" with the code...

your_git_directory% git clone https://github.com/numenta/htm.java.git

C. Proceed here to download the latest tagged release (or older if you like)

The instructions on the above link (Fork the htm.java repo) provide detail about how to fork github repos...

In addition: a video is provided that explains Numenta's contributor rules and plenty of helpful tips on using git and other commands.


Does My Code Work?

After download by clone or fork, execute a quick sanity check by running all the tests from within the /<path to git repo>/htm.java

gradle check  # Executes the tests and runs the benchmarks

--or--

gradle -Pskipbench check  # Executes the tests w/o running the benchmarks  (Faster! **Recommended**)

Note: Info on installing gradle can be found on the wiki (look at #3.) here

Linux Gradle Issues? see the wiki here.

A wiki with full build instructions for building HTM.java is available here: Build Instructions


For Developers: Usage & Project Integration

(use this stuff within my own stuff)

NOTE: The Java version of NuPIC requires no installation - just USE it!

Simple Build: For convenience, the libs directory contains all the dependency jars for those who just want to have a simple build.

...

Build Managers: Binary distributions may be included in your project using Gradle or Maven

Gradle:  

dependencies {
    compile group: 'org.numenta', name: 'htm.java', version:'0.6.12'
}
Maven:  

<dependency>
    <groupId>org.numenta</groupId>
    <artifactId>htm.java</artifactId>
    <version>0.6.12</version>
</dependency>
How to get the latest SNAPSHOT build: (None yet for newest build...)

<dependency>
    <groupId>org.numenta</groupId>
    <artifactId>htm.java</artifactId>
    <version>0.6.13-SNAPSHOT</version>
</dependency>

You also may need to include a repositories entry:

<repository>
    <id>oss-sonatype</id>
    <name>oss-sonatype</name>
    <url>https://oss.sonatype.org/content/repositories/snapshots/</url>
    <snapshots>
        <enabled>true</enabled>
    </snapshots>
</repository>

Coding With htm.java

The easiest way to use the code is to use htm.java's Network API

Please refer to the Network API Quick Start Guide to get an idea of how to quickly and simply create your own network.

...and for more in-depth answers see:

NAPI Overview


Development Environment Installation

The following instructions are for setting up a development environment.

NOTE: For simple intructions see: INSTALL.txt

Eclipse Environment Setup Wiki

An Eclipse IDE .project and .classpath file are provided so that the cloned project can be easily set up inside of Eclipse. For the time being, the Eclipse IDE is the only "pre-made" project configuration.

In addition, there are "launch configurations" for all of the tests and any runnable entities off of the "htm.java" main directory. These may be run directly in Eclipse by right-clicking them and choosing "run".


The "Fancy Stuff" - Clustering and Parallelization

Some NuPIC Community members have created interesting projects which use htm.java's Network API to run Engine-Type clusters of multiple networks using streaming distributed libraries like Akka and Flink. See them here:


To Follow Our Updates On Twitter