OXlearn - a Matlab-based Neural Network Simulator

OXlearn is a free neural network simulation software that enables you to build, train, test and analyse connectionist neural network models. Because OXlearn is implemented as a Matlab toolbox you can run it on all operation systems (Windows, Linux, MAC, etc.), provided you have a recent version of MATLAB installed (R2006b or later). This also has the additional advantage that it makes it easy to ‘look under the hood’ so you can inspect the calculations performed by the program or adapt the program to your specific needs. A compiled version (172Mb) that does not require access to Matlab is also available, currently for Windows XP only. In its current version OXlearn supports multi-layer feed-forward networks and simple recurrent networks. For more detailed information please consult the OXlearn manual.

Design:

The main purpose of OXlearn is as an educational tool that provides a quick and easy start to neural network modeling. OXlearn supports the full functionality of the well known, but now out-dated, t-learn simulator with the additional advantage of providing an extended, user friendly graphical interface – no programming is needed.

Additionally, however, OXlearn is set up for maximum transparency in terms of access to simulation parameters and data. This makes it easy to inspect and/or manipulate these values with the provided interactive tools or, alternatively, with native Matlab commands and components, e.g., to perform additional analyses or to create training corpora. More proficient users can also inspect or adjust the underlying functions, e.g., to implement custom learning algorithms, architectures, etc.

One design principle of OXlearn has been to combine ease of use with maximal transparency to facilitate extension of its existing functionality. Because of its seamless integration with the Matlab environment, OXlearn can therefore also be used as a flexible research tool.

Download :

version

requirements

instructions

OXlearn 1.0 (Matlab Toolbox, ~2MB)

all OS, but requires Matlab 7.3 (R2006b to R2010b)

Readme.txt

OXlearn 1.1 (Matlab Toolbox, ~2MB)

all OS, compatible with Matlab 2011b

Readme.txt

OXlearn 1.0 (full standalone executable, ~170MB)
OXlearn 1.0 (standalone executable, 2MB)

Windows only (tested for XP)
The full version includes Matlab Runtime Component 7.7. (MRCinstaller.exe)
See readme.txt for clarification.

Readme.txt

OXlearn manual (pdf)

 

 

Although OXlearn has not been thoroughly tested on all possible hardware configurations, it has been used in three Masters level courses and several research projects without substantial problems. Please let us know (email to: ) about your experiences with OXlearn (praise, wish lists, bug reports), as your feedback will motivate us to continue developing the software.

Terms of use:

All versions of OXlearn may be distributed and/or modified freely for personal or academic use, copyright remains with the author Nicolas Ruh. If you are using OXlearn for research purposes, please include a reference to this paper: Ruh, N. and Westermann, G. (in press). OXlearn: a new MATLAB based simulation tool for connectionist models. Behavior Research Methods.

Development of Oxlearn was supported by the ESRC.