dohoang1102 / weighted-max-cut Goto Github PK

Weighted Max-Cut

This is a document that answers the questions in the assignment.

Question 1:

    I've implemented the three functions and they get results that seem
    to sensible in comparision to each other.
    
    The results are in logFile.txt, and the implementations are in the
    .m-files that template assumed them to be in.

	I tested the files (automatically) using runAll.m which looks
	for the data files in the data folder, runAll used template.m.
        
Question 2:
   
	I didn't use the template.m to test for the two files.
	(I didn't understand why hier_sdp was commented out in template.m)
	
	Instead I ran them manually and here are the results.
	
	small_W.txt:
		gw_assign:	c=52	(This will change since it uses random)
		hier_sdp:	c=31
		
	pw01_100.7:
		gw_assign:	c=1984	(This will change since it uses random)
		hier_sdp:	c=1990
		
	I used method='average' in linkage() function.
	
	Hierarchial clustering seem to do as good as random-vector
	method for the moderate graph, but do quite bad for the
	small graph.
	
	One should however not draw statistical conclusions based on
	single samples!