Nightly Builds available

2007-03-14T21:48:07Z

In order to permit anyone interested to try the last build of the engine, starting from today we'll publish the nightly builds of the core engine.
Check periodically http://sourceforge.net/project/showfiles.php?group_id=22635&package_id=133185 in order to see if a new build is available.

Joone 2.0.0 RC1 released

2007-01-18T23:55:19Z

Joone is a FREE neural net framework to create, train and test artificial neural networks. The aim is to create a powerful environment both for enthusiastic and professional users, based on the newest Java technologies.

Joone's neural networks can be built on a local machine, be trained on a distributed environment and run on whatever device.

Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine.

After a lot of months of hard work, the new Joone 2.0 is ready. This is a release candidate, so anyone is invited to try it and help us to discover any problem.
Some of the new features:

- The core engine has been refactored by adding the single-thread mode. The resulting performances are improved by 50%, with an enhanced reliability and scalability.
- A set of tools were added, in order to hide the complexity of the API, and make it very simple to work with the core engine. Now you can build, train and interrogate a neural network by writing only 3 rows of code!
- A new Image I/O component has been added, in order read and write directly from/to image files (jpeg and gif format supported)
- Added a new SoftMax Layer to build neural networks for resolving 1 of C classification problems
- Improved the usability of the property panels, by adopting a new file chooser panel, and a visual calendar panel in order to make it very simple to insert date fields
- Several bugs were fixed and the documentation has been updated with the new features

download the last version from
http://www.jooneworld.com/download.html

JoonePad 0.5.0 released

2006-05-08T22:25:51Z

We're proud to announce the version 0.5.0 of JoonePad, the new visual editor of Joone that will substitute the actual JoonEdit.
The actual implemented feaures permit to:
* Build a new neural network by using a simple wizard
* Load a network exported (NOT saved) by the old JoonEdit application
* Display the visual representation of the network
* Edit all the properties of the Neural Network (like in the old Control Panel)
* Run/Stop/Continue the network training
* Export the neural network either in binary or XML format
* Generate the java source code needed to reproduce the edited network within a custom java program
* Select any item of the TODO-list in order to highlight the corresponding component containing the error

The new editor adopts a completely different approach to the definition of a neural network, because now it's available a wizard that drives the user step-by-step, in order to help him to build a new neural network from scratch.
Now you can open an exported neural network (from the old JoonEdit), and obtain the corresponding graphic representation in order to edit it.
Moreover, the new editor is built as an MDI application, so you can load and work on multiple networks at a time.

It is built by using the SwiXAT framework (http://www.swixat.org, another open source project maintained by P.Marrone, the author of Joone) and is just a beta release, as a lot of work has to be done.
We hope you'll appreciate the new approach.

You can launch it fron this URL: http://www.jooneworld.com/joonepad/ws/JoonePad.jnlp
(you need WebStart). If you want, you can also post your comments and suggestions at http://www.jooneworld.com/wiki/tiki-view_forum.php?forumId=2

Thanks
Paolo Marrone

JoonePad 0.1.0 released

2005-11-24T11:28:31Z

We're proud to announce the first alpha version of JoonePad, the new visual editor of Joone that will substitute the actual JoonEdit.
The actual implemented feaures permit to:
* Build a new neural network by using a simple wizard
* Load a network exported (NOT saved) by the old JoonEdit application
* Display the visual representation of the network
* Edit all the properties of the Neural Network (like in the old Control Panel)
* Run/Stop/Continue the network training
* Export the neural network either in binary or XML format
* Generate the java source code needed to reproduce the edited network within a custom java program

The new editor adopts a completely different approach to the definition of a neural network, because now it's available a wizard that drives the user step-by-step, in order to help him to build a new neural network from scratch.
Now you can open an exported neural network (from the old JoonEdit), and obtain the corresponding graphic representation.
Moreover, the new editor is built as an MDI application, so you can load and work on multiple networks at a time.

It is built by using the SwiXAT framework (http://www.swixat.org, another open source project maintained by P.Marrone, the author of Joone) and is just a pre-alpha release, as a lot of work has to be done (for example, at this stage the graph is used only to display the neural network, because the editing isn't implemented).
I hope you'll appreciate the new approach.

Joone 1.2.1 released

2005-09-09T19:47:24Z

Joone's neural networks can be built on a local machine, be trained on a distributed environment and run on whatever device.

Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine.

This is mainly bug fix release. Some of the new features:

* Added the support for the Groovy scripting language
* Added the LogarithmicPlugin to apply a logarithmic transformation (base e) to input data
* Added the save as XML in the GUI Editor

Several bugs were fixed:

* Fixed a problem that prevented the SangerSynapse from learning when in training mode
* Now the inspection panel doesn't show the biases for Layers for which this doesn't make sense
* Fixed the lacking of the first column when copied in the clipboard the inspected values

Read all the changes at
http://www.jooneworld.com/changes.html

and download the last version from
http://www.jooneworld.com/download.html

Joone 1.2.0 released

2005-02-04T18:32:17Z

Joone's neural networks can be built on a local machine, be trained on a distributed environment and run on whatever device.

Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine.

The version 1.2.0 adds many new features, among which the Radial Basis Function, a new Input Connector useful to share the same data source between several layers, a Weights' Initializing mechanism, etc.
Moreover, a lot of bugs have been fixed and the documentation has been improved.
Go to http://www.jooneworld.com/changes.html to see all the changes made in this release.

Joone DTE 0.8.0 released

2004-09-20T15:45:43Z

Joone's neural networks can be built visually on a local machine, be trained on a distributed environment and run on whatever device. Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine.

The new version of the Distributed Training Environment offers the following features:
- The training process scale almost linearly adding more machines
- Possibility to add or remove machines dynamically during the training process
- The overall process controlled by XML parameters
- Latest SUN's Jini 2.0 compliant
- Easily expandible with new features/algorithms/control modules, etc.

Read more about it at http://www.jooneworld.com/docs/dte.html

You can download the latest release here:
https://sourceforge.net/project/showfiles.php?group_id=22635&package_id=36022

Joone 1.1.1 released

2004-09-14T08:18:08Z

Joone is a FREE neural net framework to create, train and test neural nets. The aim is to create a powerful environment both for enthusiastic and professional users, based on the newest Java technologies.

Joone's neural networks can be built on a local machine, be trained on a distributed environment and run on whatever device.

Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine.

The version 1.1.1 has been released, and it is a minor bug fix release. Go to http://www.jooneworld.com/changes.html to see the changes made in this release.

Joone neural framework 1.1.0 released

2004-08-18T20:12:21Z

Joone is composed by a central engine that is the fulcrum of all applications that already exist or will be developed with Joone. Joone's neural networks can be built on a local machine, be trained on a distributed environment and run on whatever device.

Everyone can write new modules to implement new algorithms or new architectures starting from the simple components distributed with the core engine. The main idea is to create the basis to promote a zillion of AI applications that revolve around the core framework.

A new version of Joone is out! Great new features in this release:

- Added the Resilient Backprop algorithm (RPROP), a really performant new training algorithm (thanks to Boris Jansen).
- Improved the core engine's speed of about 35-40%, thanks to a deep review of the threads synchronization mechanism.
An example: On an Athlon 2200Mhz the XORMemory example took 4.4 sec to execute 10,000 epochs with the old 1.0.1 version; now it executes the same elaboration in only 2.4 seconds!
Now the Joone's performances are nearly comparable to a C/C++ framework!
- Added the setting of a http proxy to access to Internet also in presence of a firewall proxy...plus several bug fixes and a lot of small enhancements.

Joone 1.0.1 released

2004-06-13T21:09:04Z

Joone is a neural network framework written in Java.
It is modular, scalable, multithreaded and extensible and is composed by a core engine and a GUI editor to build, train and run neural networks. Can be extended writing new modules to implement new algorithms or new architectures starting from simple base components.
The aim is to build a professional and innovative framework to easily create, train and distribute neural networks that can run on any environment and device.

This is the 1.0.1 release. It contains mainly bug fixes. Also the documentation has been enhanced by adding more examples into the Joone Complete Guide pdf file.

Recent posts to news

Nightly Builds available

Joone 2.0.0 RC1 released

JoonePad 0.5.0 released

JoonePad 0.1.0 released

Joone 1.2.1 released

Joone 1.2.0 released

Joone DTE 0.8.0 released

Joone 1.1.1 released

Joone neural framework 1.1.0 released

Joone 1.0.1 released