1)  Create a 'sparse' directory in weka/classifiers and put SVMlight.java there.2)  Look at the directories for SVMlight binaries and temporary files which are defined on lines 65 and 67 in SVMlight.java.  Change these to appropriate directories in your local filesystem. 3)  Compile SVMlight.java (javac SVMlight.java).  Make sure you have ./, ../, and ../.. in your CLASSPATH (or compile using "javac -classpath "./;../;../..;../../..").If you are using Weka 3.4 or later, make Classifier the parent class instead of DistributionClassifier. 4) Put svm_classify_std.c into the SVMlight directory; replace all occurrences ofsvm_classify with svm_classify_std in SVMlight's Makefile, and compile the svm_classify_std binary by running make (the only difference between the two is output that is a bit easier to parse from Java). 5)  If you are using the Weka Experimenter GUI, add weka.classifiers.sparse.SVMlight as one of Classifier's and DistributionClassifier's in weka/gui/GenericObjectEditor.props (just Classifier's for Weka 3.4 and later).