Suppose you have a huge data set. You divide that data set into N number of blocks. Each block is then clustered into M blocks (using any clustering algorithms like K means). Now you have N local clustering solutions. Are there any methods to combine these local clustering solutions to get a globaly optimal solution?