ladoscope: tools for Logical Analysis of Data
ladoscope is a free software, which provides a bunch
of handy programs to play with the Logical Analysis of Data (LAD)
methodology. All the inputs and outputs are simple plain-text files,
so you can do whatever you want with them!
All the main features are reliable, but use the ones tagged as
-help option to know them) with care... most of
them have not been tested thoroughly, if at all. Note
ladoscope is not developed any more, but its
successor shall come some day.
There is a specific
documentation that describes the options of each program
ladoscope uses only plain-text file for inputs and
outputs; the programs are capable of reading every thing they produce
(as long as it is relevant: don't ask
read a model...). If you want to create modify or use your own way
instances, models or cutpoinset files, etc., you may want to take a
look at this sketch of
ladoscope is free software, distributed under
the GNU General Public
License (see the LICENSE file that you should get with the
programs and the
-about option for details). The binaries
are available for Linux
and for Windows
The Linux versions are the "originals", the ones I personally use. Hence, they are the most tested and the safest ones. The Windows versions are provided for convenience; they are actually only byte-code and require the ocaml byte-code interpreter and dll files (it is better to get them from the Objective Caml page but, just in case, here they are).
The sources are written
ladoscope was not started has a professional
software, but it became soon bigger than was first thought. Hence, the
sources are kind of a mess, they are inadequately documented
(especially for a third party), and tricky to compile. But they exist,
and are available.
Some handy Perl scripts are also available in a zip archive:
set-variable-names.pl: changes variable names (e.g. x0, x1, etc) to whatever you want.
monomes-of-cutpointsets.pl: displays the list of all the monomes ("x<...") defined by a cutpointset.
monomes-of-model.pl: displays the list of all the monomes ("x<...") used in a model.
set-hr.pl: changes the hr-values of patterns.
lin-approx.pl: computes the best linear-approximation of the sum of the patterns of a model.
eval-lin-approx.pl: computes the value of a function provided by lin-approx.pl for some observations.
n-folding.pl: splits a dataset in n folds, and use 1 of them as test set, the other n-1 as training sets. The number of positive (negative) observations is approximately the same in every fold.
r-sampling.pl: splits a dataset into a training set and a dataset. r is the proportion of positive (negative) observations put in the training set.
unsort.pl: mixes up the lines of a file, or simply prints them in reverse order.
Those scripts are self-documented. They are intended to work on
"standard" inputs only, and may not be capable of reading the complete
exact file formats
These scripts are the very first things I have ever written in Perl, so experts may find the style of these scripts really poor and not very perlish ;o) - but they work just fine. I owe a big thank you to my Perl guru François Gannaz.