Java bindings to FANN, the Fast Artificial Neural Network C library.

Use FannJ if you have an existing ANN from the FANN project that you would like to access from Java. There are several GUI tools that will help you create and train an ANN.

Before using FannJ, you must build and install the FANN C library. FannJ has been tested on FANN 2.2.0. See the FANN site for instructions and help: http://leenissen.dk/fann

Fann fann = new Fann( "/path/to/file" );
float[] inputs = new float[]{ -1, 1 };
float[] outputs = fann.run( inputs );
fann.close();

FANN - Does all the work.

JNA - Provides the native access to FANN.

This project is now in the Maven Central Repository.

<dependencies>
    <dependency>
        <groupId>com.googlecode.fannj</groupId>
        <artifactId>fannj</artifactId>
        <version>0.7</version>
    </dependency>
</dependencies>

JNA provides the binding from Java to the FANN C library via JNI. You must set the jna.library.path system property to the path to the FANN Library. This property is similar to java.library.path but only applies to libraries loaded by JNA. You should also change the appropriate library access environment variable before launching the VM. This is PATH on Windows, LD_LIBRARY_PATH on Linux, and DYLD_LIBRARY_PATH on OSX.

Linux something like:

LD_LIBRARY_PATH=/usr/local/lib

java -Djna.library.path=/usr/local/lib -cp fannj-0.7.jar:jna-3.2.2.jar YourClass

Update: the maven descriptor (pom.xml) now contains a section (atm only for Linux) that builds FANN from sources and packs the native library into a jar file called fannj-<version>-linux-x86_64.jar. If you put that jar onto the classpath, the JNA libs should be able to extract and load the contained libs all by themselves and without additional configuration.