A boring computing-page by Axeloide

Programs (everything is Freeware written by myself)

Join 1.1 for DOS written in C

This program takes several inputfiles with two TAB or SPACE delimited columns of numbers wherein the first column of all inputfiles must be identical and generates a single outputfile containing the first column of the inputfiles followed by all the second columns.

You can use this program to join multiple files as generated by spectrometers, which typically contain in the first column the wavelength-numbers and in the second column the absorbtion-values. Instead of importing every file into your spreadsheet or graphics-program, you can join them first using Join into a single file and afterwards import this Join-outputfile into your program.

Join will check inputfiles as follows:

Must have two columns separated by Tab and/or by Space
Must have same number of values. This is: no missing values and same number of rows
No letters, except e used in the scientific notation (eg.: 1.234E7), are allowed.
The first column must be identical in all inputfiles.

If any of theese is not given, Join will terminate execution.

Join has the following options:

Transpose: Columns in inputfiles become Rows in outputfile
Sort inputfiles by date/time or by extension. Default sorting is by filename.

SYNTAX
        join filemask [/?] [/Oname] [/Xtitle] [/St] [/Se] [/T]

filemask   :    Filenames matching this expression will be joined.
                Wildcards are '*'  and '?'
                eg: join *.dat
                    join sample??.*

/h OR /?   :    Display this info

/Oname     :    name is filename of outputfile
                eg: join *.dat   /oresult.txt
                will join all files matching *.dat into result.txt
                If not specified, outputfilename will be joined.txt

/Xtitle    :    title is title of X-column (or row if transposed) in outputfile

/St        :    Sort columns by file-timestamp
/Se        :    Sort columns by extension
                Default is sorting by filename

/T         :    Transpose in outputfile: columns become rows.

Download Join11 (ZIP-File: 36.679 bytes).

PovCol 1.0 for DOS written in Pascal

This program takes the X, Y and Z coordinates from a TAB and/or SPACE delimeted ASCII-file and generates a POV-file, which compiled with the POV-Raytracer, produces a 3D-Fotorealistic picture, displaying the data from the ASCII-file as little spheres in 3D-space.

See an example-picture here.

POV-Ray can compile several images from diferent angles, so you can merge all the resulting TGA-pictures into a FLI-movie where 3D-space is rotated. You can translate TGA-files into a single FLI-file using the program VFD(tm) Video for DOS written by Bob Williamson.

See an example-movie here. It's a 1.23Mb AVI-file. Very bad quality. I don't recomend you to download this movie if you are conected via modem!

Download PovCol1.zip (ZIP-file: 64.329 bytes). It contains the executable, the Turbo Pascal source-code and an example ini-file needed by Pov-Ray to compile the scene-file (*.pov) into several TGA-files.

TrainNN for whatever written in C

TrainNN will generate a Neural-Network (NN) and train it with user-provided input-output-pairs. The structure of the NN is of an multilayer Perceptron (or feed-foward-network). The user can decide how many levels and neurons per level the NN will have. TrainNN uses evolution-strategy to train the NN. When the training finishes, it saves the calculated weights of the NN into a file, which can be read by the program TestNN.

How it works

TrainNN reads an Option-File like the following:

NNstruct=       NNstruct.dat
WeightsIn=      NULL
WeightsOut=     w_out.dat
Training=       train.dat
Report=         report1.txt

EveryGenerations= 20
MaxGenerations= 10000
Criterium=      0.00000000001
Parents=        1
Descendents=    20

Format of the Option-File

NNstruct=
In this example TrainNN will read a Neural-Network-Structure-File called NNstruct.dat like this:
```
36
4

20
10
10
1
```
Format of the NN-Structure-File:
- The first number is the number of inputs to the NN.
- The second describes how many levels the NN has.
- The following numbers define how many neurons each level has, beeing the first number the number of neurons in the first level. The last number is the number of neurons in the outputlevel. Recall that NN have as many outputs as the last level.
So in this case TrainNN will build a 4-level Perceptron with 36 inputs and 1 output.
WeightsIn=
File with initial weights of the NN.
Write NULL to initilize the NN with all weights set to 1.0.
Or write the name of a weights-file as generated by TrainNN. See option WeightsOut=.
WeightsOut=
File where TrainNN will store the weights of the NN after training it.
Training=
Name of the Training-File TrainNN will use to train the NN.
Example of a Training-File:
```
36
1

0 0 0 0 0 1
0 0 0 0 1 0
0 0 0 1 0 0
0 0 1 0 0 0
0 1 0 0 0 0
1 0 0 0 0 0

1

1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1

0
```
Format of the Training-File:
- First number is number of inputs to the NN.
- Second number is number of ouputs from the NN.
- Following numbers are input-output-pairs the NN has to learn.
This Training-File will try to teach the NN to assign the output 1 to a slash and assign the output 0 to a back-slash.
Report=
Name of the Report-File that documents the training process.
EveryGenerations=
TrainNN will refresh the display every X training-steps. Information about the quality etc... is displayed.
MaxGenerations=
Maximal number of training-steps to perform.
Criterium=
When the quality-function returns a value smaler than this, then the training will be finished.
Parents=
Parameter for the evolution-strategy: Number of Parents to select.
Descendents=
Parameter for the evolution-strategy: Number of descendents to produce in each generation.

.