SourceForge.net Logo Download
Support This Project
[Home]
[About AINN]
[Download]
[Links]
[Documentation]
[Neural Entity]

 

  1. Mutanor: a small game using the NN

This small sample is available under the following form:

  1. Neural Entity Package

The new package is developed with the VC2005 (beta 2). This IDE and compiler are available at microsoft's lab site The IDE is quite good and the code produced is reliable. I really do recommend the product.

In this new design, I take the advantage of the dll's - more convenient to update and offers a better modularity (under linux/unix, people can also make use of the so objects).. I also take advantages of the threads, critical sections, events and internet connection from Microsoft (I do apology to the unix/linux pro's but it was so easy with Microsoft that I could not resist).


All the sources and samples are developed using the VC8++ express - the old-way still makes use of the VC6 and VC7

  • Under windows: the style of compilation is pure ANSI-C++ (__STDC__ with no ms extension in use) also the scope of variable declaration in the for loop are conform to the standard.
  • Under Linux: the compilation make usage of the 'ansi' flag.
  • Warning: the current library make use of the STL. The end-users willing to use the AINN with VC6 should use the STL-Port or any other good template library but NOT the one provided by Microsoft as it is not standard - Using VC7.x or VC8.x should not make any problem as for these versions Microsoft has corrected the issues. I haven't encountered any issue under Linux but make sure you have a correct standard library.

 If a file is missing, it means it is under way and will shortly be available.

  1. AINN-Old-WayBase library

This is the old-way versus my way regarding the neural network. This library has the definition of three neural networks: the adaline, the kohonen and the bakpropagation. The library will merely be maintained and is present just to illustrate the difference between an old-way (applied mathematical one) and my own perception of the solution (informatics one).

The library is a collection of object representing a neuron (Soma, dendrites and Axon) and the base structure of a network organized in different layers (Cfr. Documentation also available to download here).

Only the sources are available and can be downloaded here.

  1. Neural Entity
  • Version 1.0 Draft 1 is a basic version with only the base neuron definition and a sample of neuron interaction showing how to create and and, or, not and eor gates (Cfr documentation). The library is subject to heavy modification - consider it as a draft and theoretical, I intend to create a less theoretical library and by the faster... But I am not there yet!
  • Version 1.0 Draft 2 is the newly approach of the neural entity philosophy. Indeed, we show how to create a simple eor gate with two neuron and no learning require at all.... Sounds like magic when I compare with the effort provided by the back propagation for example. The C++ code is in test1-bio.cpp file. Just enjoy how easy and intuitive it is.... The next version will start to implement an automation flow of the Depolarization() and the beginning of the learning ratios.
  • Version 1.0 Draft 4 is the newly approach of the neural entity philosophy. This draft version present a new type of neuron interconnection...
  • Version 1.0 Draft 5 has some basic adjustment /learning capabilities with a small sample.
  1. Small utilities