1. Type "make" in this directory to create executable files:
	nksvm: MultiLinear SVM
	lsvm: Linear SVM
	p2svm: Degree-2 Expansion SVM
	featurewise: featurewisely scale the input data
	featurewise2: featurewisely scale both the training and testing data
	instancewise: instancewisely normalize the input data

2. Go to experiment-py/ directory

3. Experiments on Scaling:
	(a) make -C ../liblinear-2.1
	(b) ./scaling.py <path_to_data>
	(c) A log file: dataname.log is created in log_files/ directory
	(e) Some additional files are also created in transformed_data/ directory
		dataname.f: the data after featurewisely scaled
		dataname.i: the data after instancewisely scaled

4. Experiments on Degree-2 Expansions & MultiLinear SVM:
	(a) ./runML.py <path_to_data>    (dataname.klog created)
	(b) ./runSML.py <path_to_data>   (dataname.sklog created)
	(c) ./runP2.py <path_to_data>    (dataname.p2log created)
	(d) Three log files: dataname.klog, dataname.sklog & dataname.p2log are created in log_files/ directory
		dataname.klog: Stores information on running MultiLinear SVM with kmeans
		dataname.sklog: Stores information on running MultiLinear SVM with spherical kmeans
		dataname.p2log: Stores information on running Degree-2 Expansions with different r ( = coef0)
	(e) Some additional files are also created in transformed_data/ directory
		dataname.f: the data after featurewisely scaled
		dataname.f.tr: roughly 3/4 of dataname.f
		dataname.f.val: roughly 1/4 of dataname.f
		(The last two files are obtained using train_test.py,
		 which assigns each instance to a validation set with probability 1/4)
