ATRE learning system is only developed for Windows platforms and is composed
by two main components: the ATRE learning engine (developed in Sicstus-Prolog
3.8.6) and the ATRE Graphical User Interface (developed in Visual C++ 6.0). As
a consequence, you can only download the ATRE learning engine or you can download
the entire system (including both the components).
Downloading the ATRE learning engine (without GUI), you need to unzip the package distribution
file into a directory and run the ATRE.EXE file as specified below:
atre learn <filedat> to perform a learning task
with ATRE using the training data file <filedat>
If you download the entire system, you need to unzip the package distribution
file into a temporary directory that will contain the installation package. Then,
you have to run the SETUP.EXE to install the software. Further details on GUI
usage are presented in the next section.
This menu allows to
select the input file for ATRE. When you select the Load data menu
you will be presented with the Open
dialog. Double-click on the chosen directory in order to view its content in the
Files box. Select the desired file and click on the OK button or press the
<enter> key in order to load it The Load training data
command allows to select the input file to train the system. The Load test
data command allows to select the input file to run the recognizer. The
Load rule data command allows to select the input rule file for the
recognizer.
2.2 The Learn process Menu
The menu manages ATRE's training process.
The View training data command shows the input file to train the
system. The View rules command shows the rule file produced by ATRE for
the current process of training. The View learn report command shows the
output file produced by ATRE for the current process of training. The
Learn parameters command allows to set the learning parameters for
training process like verbosity,general,best lef and ps lef. The
Verbosity parameter defines which informations save in the report file,
like objects, examples, clause selection, generated clauses and clause consistent
and range-restricted. The General parameter allows to set the maximum
number of clauses to specialize in a single specialization step (MaxStar), the
least number of consistent clauses to generate before to select the best
(Consistent), the maximum number of literals for the specialization of a clause
(MaxPs) and the check-box Recursion that enable the search for recursive
definitions. The Best lef and the Ps lef parameters allow to
define the preference criterions, the first for the selection of the final rule
and the second for the specialization phase. The Start learning command
starts the training process for ATRE.
2.3 The Recognition process Menu
The menu manages ATRE's recognition
process. The View test data command shows the input file to test the
system. The View rules command shows the rule file used for input in the
recognition process. The View recognition report command shows the output
file produced by ATRE for the current process of testing. The Start
testing command starts the recognition process for ATRE.
A session of
training with ATRE could be illustrated as: 1. Choose the input file
(*.dat) to be trained; 2. Start learning process clicking on the Start
learning command. The system produce the output file (*.out) with the
description of every steps of theory's generation and the output rule file
(*.rul) that describes the generated theory. The output files are called the
same as the input file.
A session of testing
with ATRE could be illustrated as: 1. Choose the input file (*.dat) to
be tested; 2. Choose the rule file (*.rul) to use as input to the
recognizer; 3. Start recognition process clicking on the Start testing
command. The system produce the output file (*.rec) with the list of
commission and omission errors for every object and for every rule. The report
file is called the same as the input files.