- 
                Notifications
    You must be signed in to change notification settings 
- Fork 0
Git clone for JavaNNS Neural Network System
License
ycollet/JavaSNNS
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
************************* JavaNNS README *************************** Contents -------- 1. Prerequisites 2. Building JavaNNS 3. Compiling Kernel interfaces 4. NetBeans 4 support 5. Known Bugs 1. Prerequisites ---------------- In order to build JavaNNS you will need: - A current Java Development Kit (JDK) (we recommend Sun's JDK) - A properly set up Ant Development Environment (http://ant.apache.org) 2. Building JavaNNS ------------------- JavaNNS actually consists of two separate modules, the Java GUI and the SNNS kernel written in C. This source code distribution of JavaNNS supplies precompiled kernels and kernel interfaces for several platforms. Setting or changing the target platform does not modify the Java source code or its compilation, but merely defines which precompiled kernel interface will be put into the final JavaNNS.jar file. The following list describes the targets which are implemented in out Ant build script and their respective actions: javanns [default] - Compiles JavaNNS and asks for the target platform, if it has not been set before. Use this target for easy debugging, as all .class files will be put into the respective source directories. jar - Compiles JavaNNS and packages all necessary components into a JAR file (dist/JavaNNS.jar) run - Compiles JavaNNS (if necessary) and executes JavaNNS settarget - Allows you to change the target platform (Windows, Linux, Mac, Solaris) clean - Deletes all intermediary .class files. Please note that the dist directory won't be removed help - Displays this help text 3. Building kernel interfaces ----------------------------- This source code distribution of JavaNNS comes equipped with precompiled kernels and kernel interfaces for - Microsoft Windows - Linux (x86) - Sun Solaris - Mac OS So if you intend to build JavaNNS for a target not listed here or modify the SNNS kernel, you will need the SNNS sources, which can be downloaded from http://www-ra.informatik.uni-tuebingen.de/downloads/SNNS Compile the kernel with an appropriate C compiler, e.g. GCC for *IX platforms or Microsoft Visual C++ for Microsoft Windows. After this has been accomplished, create a shared library / DLL by compiling javanns_KernelInterface.c from the KernelInterface directory and linking it together with libfunc and libkernel (which were created during compilation of the kernel) into a library. For this to work, your compiler must be supplied with: - the path to the include directory of your JDK which contains JNI header files - and the path to the SNNS kernel source-directory (e.g. SNNS/kernel/sources) Then overwrite the old library in the directory KernelInterface/precompiled/[platform] with the newly created one. Unfortunately, there is no easier or quicker approach available at the moment. Please note that in order to really use this updated kernel interface, you must copy it to the location specified in 'JavaNNS.properties', located in your home directory. 4. NetBeans 4 support --------------------- We do not intend to encourage or discourage the use of a certain IDE to perform work on JavaNNS. Anyway, Sun's NetBeans 4 allows for easy integration of Ant based projects. We recommend to map NetBeans tasks on Ant targets as follows: Build Projects -> javanns Clean Project -> clean Run Project -> run Debug Project -> debug 5. Known Bugs ------------- - The "Patterns" tab in the Control Panel does not work as intended. Trying to add patterns or pattern sets for a network which does not contain any patterns so far, will lead to a fatal crash. Please circumvent this bug by loading patterns from a .pat file. With at least one pattern already loaded, this bug does not occur.
About
Git clone for JavaNNS Neural Network System
Resources
License
Stars
Watchers
Forks
Releases
No releases published
              Packages 0
        No packages published