readme.txt

ANN 1.0 人工神经网络程序, Artificial Neural Network Training Program ANN.EXE v1.10,

                Artificial Neural Network Training Program
                ANN.EXE v1.10, Copyright (c) Sommnath Kundu
                            1994-1996

DESCRIPTION:
-----------
This artificial neural network program (ann.exe) was developed in the
Department of Chemical Engineering, Calcutta University, India, as a
part of a research project on the application of ANN in the field of
Chemical Engineering. It may be quite helpfull to those who are doing
research or interested in this field. It is a general purpose ANN
training program. You can use input/output data for any system to
construct and train the network. To effectively simulate the system
you must include all the key parameters of the system as input or
output of the net. This program is menu driven and will ask you for
required inputs. All the options will be saved in a configurayion file,
so you can use it later as a command line parameter. Followings are
some of the usefull features of the program,

    1.) Any size of the network can be constructed as permitted by DOS.

    2.) Different training algorithms are available, e.g., steepest
        descent, conjugate gradient, DFP, BFGS and Maquardt-Levenberg.

    3.) A heuristic search algorithm partially based on quadratic
        interpolation, developed by the author, is used to find out
        optimum stepsize for each iteration. This algorithm increases
        training speed and accuracy many times.

    4.) Different mode of training can be selected, e.g., pattern,
        moving window and block mode of training.

    5.) Activation function can be selected.

    6.) Input and output data to the net can be scaled automatically.
        So you can instantly compare and verify the trained net for the
        test data set.

    7.) Various intermediate results during training can be stored in
        the data files to facilitate statistical analysis.

There is a single executable file, ann.exe. You can run it interactively
or non-interectively by giving configuration file name as a command-line
parameter. All the data files used by the program are formatted as a text
file, so you can use externally created data file, but the format should
be the same. 

SAMPLE DATA:
-----------
A set of sample data files are included for a model of  N-number
of equal volume CSTR (constant stirred tank reactor) connected
in a series having concentration CA0 (of A). At time t0 some amount
solution A of concentration CA is fed into 1st tank. The concentration
of A in the last tank is calculated as a function of time and number
of tanks present in the series. The concentrations at different time
interval are used as inputs and number of tanks as outputs to train
the net. The data files included with this program are as folows:

        CSTR_N1.DAT  ==>  Training data set for CSTR model
        CSTR_N2.DAT  ==>  Test data set for CSTR model
        W0           ==>  Initial weights
        W1           ==>  Final weights (to be calculated)
        TRNG.ITE     ==>  Iteration record file for training data
        TEST.ITR     ==>  Iteration record file for test data
        CSTR.DTL     ==>  File to save details calculation
        CSTR.CFG     ==>  Net configuration file

You can train the net for this sample data by giving net configuration
file (CSTR.CFG) when asked by ANN.EXE or giving it as a command-line
parameter, e.g., ann cstr.cfg.

All the above data files will be generated for your data file.

COPYRIGHT:
---------
ANN v1.10 can be freely distributed, but it should not be used or
distributed for any commercial purposeS without prior permission
of the author.

In case of any problem or question you can contact the author in the
following addresses.

Somnath kundu
Office:    [e-mail]
Residence: 44/A Madhu Roy Lane,
           Calcutta - 700 006, India.