The high-level interface to the library is using ASCII-text files. The advantages of this approach is that text editors, filters, or tools like vi, sed, grep or awk can be used to maintain and manipulate the files. There are also UNIX-programs working on text files for visualization of the data; in the project the program gnuplot was used.
This makes it easier for an user to view data input files as well as results written by the system. The exchange of files between different platforms is straight forward when using ASCII-text files. It is possible to train the system on a fast UNIX-workstation and then copy the network description file to a PC and use the trained network without conversion.
The main disadvantage of using text files is that they are larger than the equivalent binary files would be. The longer time needed to read the file from disk is not significant in comparison to the training time of the network; particularly by using the class CData where the file is only accessed once.
To archive larger data files or network description files, all platforms offer tools to compress the text files; on the UNIX system compress and on the PC pkzip were used.