% % *************************** % TOOLDIAG vers. 1.4 % *************************** % % A experimental pattern recognition package % % Copyright (C) 1992, 1993 Thomas W. Rauber % Universidade Nova de Lisboa & UNINOVA % Intelligent Robotics Center % E-Mail: tr@fct.unl.pt % % % INTRODUCTION % ------------ % % TOOLDIAG is a experimental package for the analysis and visualization % of sensorial data. It permits the selection of features for a supervised % learning task, the error estimation of a classifier based on continuous % multidimensional feature vectors and the visualization of the data. % Furthermore it contains the Q* learning algorithm which is able to generate % prototypes from the training set. The Q* module has the same functionality % as the LVQ package described below. % Its main purpose is to give the researcher in the field a feeling for % the usefulness of sensorial data for classification purposes. % The visualization part of the program is executed by the program GNUPLOT % which runs on a variety of platforms (DOS, UNIX etc.). You should therefore % posses a copy of the public domain program GNUPLOT. If you do not have % GNUPLOT see the file 'other.systems' how to obtain a free copy of it. % The 'TOOLDIAG' package has also an interface to the classification % program 'LVQ_PAK' of T. Kohonen of Helsinki University and his programming % team. See also the file 'other.systems' how to get a copy. % Also the Stuttgart Neural Network Simulator (SNNS) can be interfaced. % It is possible to generate standard network definition files and pattern % files from the input files. Therefore TOOLDIAG can be considered as a % preprocessor for feature selection (use only the most important data for % the training of the network). % % HOW TO OBTAIN TOOLDIAG % ---------------------- % % The TOOLDIAG software package can be obtained via anonymous ftp. % Server: ftp.fct.unl.pt % Directory: pub/di/packages % Files: tooldiag.README % tooldiag.tar.Z %