What is CLOP?
CLOP stands for Challenge Learning Object Package. It is
an object-oriented Matlab(R) Machine Learning package. CLOP is based on
the Spider
developed at the Max Planck Institute for Biological Cybernetics and
integrates software from several sources, see the credits. It was
developed to support challenges
in Machine Learning, and it includes the best working methods from
these challenges, in addition to the Spider library. You can also
access from CLOP Weka and R functions.
CLOP and the Spider are built on two simple abstractions: data
and algorithm. Once you load some data matrix X and target
vector Y and create a data object: > training_data =
data(X, Y);
You just need to instanciate an algorithm, say a Support Vector
Classifier (svc): > my_svc = svc;
Then you train it by calling the method train (in Matlab,
functions, which have an object as their first argument are methods of
that object): [training_resu,
trained_svc] = train(my_svc, training_data);
The object resu contained the predictions on training data. The trained
model may then be tested with the method test on test_data test_resu =
test(trained_svc, test_data);
Compound models can be built by chaining algorithms, including
preprocessing, predictors, and postprocessing and/or building ensembles
of models voting towards the final decision. The resulting compound
model is then trained and
tested by calling train and test; it knows how to train and test itself
by
calling the train and test methods of its components.
Download CLOP
Before you download CLOP, please make sure you read the license agreement
and the disclaimer.
There are several versions of CLOP. We recommend you download the
last one:
Book on feature selection version: Featbook
version (April 2005)
Installation instructions
==> Windows users will just have to run a script to set the Matlab
path properly to use most functions.
==> Unix users will have to compile the LibSVM package if they want
to use support vector machines. Please use the latest
Makefile.
==> All users will have to install R to use random forests (RF and
RFFS). Make sure you remove and file named
Clop/challenge_objects/packages/Rlink/__Rpath. When you first start RF
or RFFS, you will be prompted for the path of the R executable.
Feature
Extraction, Foundations and Applications, Isabelle
Guyon, Steve Gunn, Masoud Nikravesh, and Lofti Zadeh, Editors.
Series Studies in Fuzziness and Soft Computing, Physica-Verlag,
Springer, 2006.