Intuitive Neural Network

Home Page

Abstract

AI methods are regularely used in competer science for they provide a different approach which is often preferable. These methods are not always easy to understand so they have to be more accurately described in order to enhance comprehensibility. This projects aims at a gathering of CAE, computer graphics and AI to provide a neural network which will be able not only to produce a solution of a problem not easily modelisable but also to exhibate the way it is "thinking"

GUI and 3D visualisation

To be intuitive INN provides a GUI which presents multiple features :
  • Easy neural network modeling
  • Wizards
  • 3D visualisation
  • Learning speed control including a step by step mode
 INN interface
INN interface
INN therfore permits to evaluate the way the neural network is learning providing graph tools and other statstics. A testing interface allows to add some noise to a sample which allows to evaluate the example teaching abilities.


Crossplatform application

INN is written in C++ with the QT framework. It uses also C++/QT libraries such as libQGLviewer for 3D interaction and visualisation. As using the QT framework and the provided tools, such as uic and qmake, INN is highly portable is its source will compile whatever the platform without changing a letter in the source code. An XML formalism has also been implemented in order to save a network teached or not.

Crossplatform application