MEHRANN
and MPE software
Mehran Hoodeh, MPE IDE Demo  
Download Demo
File format: .mp4
File size: 9 MB
 
 
 
  Mehran Hoodeh, MEHRANN Logo  
     
  Short story:
MEHRANN is the name of my Back-propagation Neural Network, and MPE is the programming environment specially developed for it.

Long story:
Artificial Intelligence course was a great part of my MSc in which we became familiar with different types of Neural Networks, but for good reasons we were focused on and practically worked with "Feed-forward Back-propagation Neural Networks" shortly known as BPNN. In this course, we needed to work with MatLab to build the Network, train it on a set of data and test it on unseen data.

This was when I thought I could develop a software in which doing the configuration of a BPNN would be easier and faster and could be used widely in different kinds of projects.

With this idea in mind, I developed MPE (MEHRANN Programming Environment) in which a BPNN can be easily built up, trained, tested, saved in a file and re-used in future as a predictor/classifier. For this I needed to develop a BPNN beside an IDE (Integrated Development Environment) for programming it.

MEHRANN which is the name of my BPNN, stands for Mega Edged Highly Reusable Artificial Neural Network.
- It is Mega Edged because you can have a network with millions of neurons/edges in it.
- It is Highly Reusable because you can use it in any kind of projects that require a BPNN for decision making, predictions and classifications.
- It is Artificial Neural Network simply because it is an Artificial Neural Network; a known tool in Artificial Intelligence applications.

MPE provides the user an IDE to write the program in a specific grammar. This IDE compiles the program, finds syntax errors and runs the program. The following shows a snap-shot of the software with a simple program loaded in it which trains a Neural Network on a set of "Heart Data" for classification of a Heart Disease.


Mehran Hoodeh, MPE IDE Snap-shot




And when the program is run it loads the data-set and starts training showing the RMS error graph while training and a graphical view of the designed network with active edges (when mouse hovers over an edge, a neuron or the bias the weight or neuron info is displayed as shown below).
To see how a program is written or opened and run, or how the Network starts training with RMS Error Graph and how the structure of the Network is visualized, you can download a Demo movie of the software showing how it works.


Mehran Hoodeh, MPE Animated Demo