Boone

Basic Object-Oriented Neural Environment

The Java framework Boone assists in developing Artificial Neural Network (ANN) applications. It follows a strict object-oriented approach focussing on the extensibility of the framework, hereby, making concessions regarding computational efficiency. However, the simple, and biologically motivated, approach to center the framework on only two main classes, namely Neuron and Link, may even be beneficial to the speed of execution. Arbitrary feed-forward or recurrent networks can be constructed, which may be trained by integrated Trainers using pattern sets, for which different file formats are supported. Lately, a package supporting Spiking Neural Networks (SNNs) has been added to the framework, which has been used and tested in a number of applications, e.g., robotic neurocontrollers or neural Go players.

Boone Features

Lean and compact Java package
Supports arbitrary network types
Feed-forward trainers based on error back-propagation
A basic Hopfield trainer
Self-organizing map (SOM) trainers
Spiking neural networks based on discrete event simulation
Filters for ANN file formats (including SNNS)
Filters for pattern file formats (including SNNS)

Binaries, Documentation, and Sample Programs

Boone reference
(please, be aware that parts of it are outdated as of now)
A small sample program
The complete package (including API)
(if you are using the software for any non-private use, it is highly appreciated, if you are citing it, note that in any case you MUST adher to the respective licenses of the included JDOM and JAMA packages)

Last modified: Nov 27 2023