We have done three experiments
	- size and empirical error tradeoff
	- speeding up kernelized k-means++
	- speeding up spectral clustering

-------------------------------------------------------------------------------------------
Compile:
	- make
Run:
	- ./exp1	to run the first experiment on twitter dataset
	- ./exp2	to run the second experiment on twitter dataset
	- ./exp3	to run the third experiment on twitter dataset

-------------------------------------------------------------------------------------------
Directory Overview:

exp*.cpp				the main cpp to run the experiments
parameters.h 			parameters of the dataset and the kernel function
parameters*.h			parameters of each experiment
tools.h
tools.cpp 				some useful tools, such as (pseudo) random number generator
kernel.h
kernel.cpp 				definition/implementation of kernel functions and metric functions
kmeans.h
kmeans.cpp 				definition/implementation of kmeans++
coreset.h
coreset.cpp 			definition/implementation of coreset
empirical_error.h
empirical_error.cpp 	calculating empirical error
