A Neural Network modeling environment
Sevenn, written entirely in Perl, uses MySQL for storage and caching. It uses SOAP to exchange data between the client and the server and parses and executes NNL (Neural Network Language) statements. There is no limit to the number of nodes, edges, networks, and functions you can store and simulate other than the amount of space available to store them on the server. All documentation and source code is included in the github repository:
http://github.com/danndalf/sevenn/
Open Source Neural Networks
- Unlimited nodes and edges
- Modeling of propagative effects
- Open Source, Free Software
- Server-side processing
Once Sevenn is installed, there is a walkthrough available in the Documentation folder of how to build a basic hot-cold switch detection network. This is done by issuing commands in a constructed syntax called NNL (and used, to my knowledge, exclusively in Sevenn) to create nodes, edges, networks, and activation functions. Basic functions such as sinusoidal thresholds are included, though new ones can easily be added as needed.
License
I wrote Sevenn as a research project at university as a proof-of-concept of connectivist neural modeling on low-end (at the time) equipment. It is licensed under the GPL version 2. For more information, see the license.txt file enclosed with the distributed archive.
Note: Sevenn is no longer actively supported.
This package is freely available as an example way of modeling neural networks, though is by no means the best available technology. If you're seriously looking to get into neural network research, there are much better tools than this freely available to you, such as:
- FANN, the Fast Artificial Neural Network library. FANN has good bindings available for most common languages.
- AI::NNEasy, if you want to stick with perl.
- Flood, an open source C++ Neural Networks library.