========================================================================
Title:	Discrete Neural Network
Version:	1.0
Author:	Sergius Adamtchouk (http://www.geocities.com/sadamv) 
========================================================================

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
		Context
1. Description
2. License agreement
3. Files description (Installation)
4. Instruction for usage

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
		1. Description
This program is intended for searching an automatic dependence one real vector on another.
Principle of working is based on the Neural Networks Model. This is very little program, but it has greate possibilities.

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
		2.  License agreement
Everybody can use this program as he wants.

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
		3. Files description (Installation)
1) NeuroNet.exe	- executable modul
2) *.dat		- examples with training data, consult or report sets.

NOTICE! The program need mfc42d.dll file.

		4. Instruction for usage
	1. General information 
The program is created in the form of a documented architecture, which allows working 
with neural networks as documents with training data. Neural networks can be stored/loaded 
on/from a disk, print out, etc. 
Each neural network is characterized by the number of layers, enters and exits.
	2. Selection of the training data
Training data are called the consistency of patterns, each of them consists of a number of 
input and output real values.  For selection of the training data they must be saved in the 
ASCII file. For each input and output values is formed a separate transforming table, 
which express the connection of this real parameter with the internal discrete value, which 
are manipulated by the neural network. Transforming tables like the training data are saved 
with the documental file. By the selection of the training data we can use transforming tables
created for the training data selected before or create new tables  Create New
Transformation Tables. 
	3. Randomization
This option reinitializes all weight coefficients into the random values while destroying
experience got before. It is used for the solution of the local minimum problem.
	4. Consulting
The option of consulting provides the final usage of experience got by the neural network.
There are two options of consulting:
Singular  all input values are entered by the user; the system finds and  shows output
values.
 Plural  the user assigns names of two ASCII files. The first one is the file of input values,
where the sets of input values. For each set the system finds output values and stores in a file,
which name is entered in the field of the output values file.