Generate lots of different synthetic data, run it with each algorithm, and compare results with respect to whichever notions of accuracy have been implemented.
It would be good to perform some experiments on exactly the same data as used in existing papers. It seems that several well-known papers use data from this site: http://lunadong.com/fusionDataSets.htm
Performance in terms of time and memory usage should also be compared.
Generate lots of different synthetic data, run it with each algorithm, and compare results with respect to whichever notions of accuracy have been implemented.
It would be good to perform some experiments on exactly the same data as used in existing papers. It seems that several well-known papers use data from this site: http://lunadong.com/fusionDataSets.htm
Performance in terms of time and memory usage should also be compared.