Project Description

Eddie's Neural Network Framework is an experimental .NET Library to make the creation, modification and storage of Neural Networks easier.

Technologies used

.NET 4.0, Enterprise Library 5.0 Optional Update 1; Entity Framework 4.1; Windows Installer Xml 3.5; Math.NET Numerics; Prism 4.1; Code Contracts; 3dTools

Remark: To compile, please also download the Enterprise Library 5.0 Optional Update 1 , Prism 4.1, Entity Framework 4.1, Windows Installer Xml 3.5; Code Contracts and Math.NET Numerics

For Details to the Enterprise Library look on MSDN in the Patterns & Practices

For Questions about the Entity Framework 4.1 check out the ADO.NET Team Blog

Roadmap

I’m doing this for the fun of it, so basically this is more an orientation point then anything else and “subject to change”…

Phase 1 – Feed Forward Neural Networks

The first step is to create the basic infrastructure – the abstract Interfaces, the plumbing which reads, writes and updates the data to the database, etc.

The second/parallel step is to implement a basic Feed Forward Neural Network. It will allow for activation functions, feedback learning and other “usual” things used in a FFNN.

Phase 2 – Expand the features of the Feed Forward Neural Network

Once the basics are covered, Phase 2 will expand on the existing infrastructure and implement a wide rand of options – different kind of activation functions, different learning strategies, etc.

Phase 3 – Implement Radial Basis Neural Network

The next phase will consist of implementing different kinds of networks such as Radial Basis Networks.

How can you help?

For now so far the Tests have been tested against SQL Server 2008 R2, testing against other types of databases would be a great help.

Last edited Nov 6, 2012 at 10:44 PM by ebralph, version 17