Package boone.io

Loading and saving Boone neural networks and patterns to files or streams.

See: Description

Interface Summary

Interface	Description
Storable	Interface with methods to load and store an object instance's data from/to XML DOM.

Class Summary

Class	Description
BooneFilter	An I/O filter for the native Boone file format with optional GZip compression.
CSVPatternFilter	The CSV pattern filter assumes that a complete pattern (input and target) is written in a single line (a record), where the values are separated by a specific character (default is the comma ',').
IOFilter	This is the base filter for loading and saving networks and patterns.
Proben1PatternFilter	A pattern filter for PROBEN1 pattern files.
SNNSNetFilter	Load and save SNNS net files.
SNNSNetFilter.ENSNNSLexicalAnalyzer	Lexical Analyser class.
SNNSNetFilter.ENSNNSParser	Parser.
SNNSPatternFilter	A filter for for SNNS pattern files.

Exception Summary

Exception	Description
BooneIOException	Abstract base class for exceptions during file handling.
BooneIOException.Parsing	Exception while reading / parsing a file.
BooneIOException.Writing	General exception during write.

Package boone.io Description

Loading and saving Boone neural networks and patterns to files or streams. Contains basic infrastructure for handling the Boone XML storage format, as well as some converters to non-native formats (such as SNNS or PROBEN1 patterns).

Some classes are for reading and writing patterns in the PROBEN1 benchmark file format. To benchmark neural network algorithms, there are a number of test patterns available with the PROBEN1 test suite; these can be read and written with the classes in this package. Please observe that, to produce sensible data, you should also observe the recommendations in the PROBEN1 paper.

There are also some classes for reading and writing SNNS training pattern files. These classes don't use the Boone I/O framework, but analyse the objects directly; this is because they need to be compatible with existing datafiles.