Graph Giant Connected Component

Question

i have a matrix of the form

a b 8.0
a d 0.1
......

where 1st column is Node A,2nd Node B and 3rd correlation coefficient i have to make a program that finds a threshold thus the connected network has a Giant Connected Component consist of 50-60% of total Network nodes. I wrote a program that using a binary search for the threshold like

if Giant Connected Component > 60% new threshold=oldthreshold + oldthreshold/2
if Giant Connected Component < 50% new threshold=oldthreshold - oldthreshold/2

The problem is that algorithm is also searching for thresholds > 1 and/or <0 .How i can handle this.Or is there any better idea how to calculate it?

You could find the minimum and maximum values in the table and use those as the starting range for binary search. — augurar, Nov 20 '16 at 07:33
Ah, I see. Your problem is that you are not correctly implementing binary search. I'll answer in more detail in a moment. — augurar, Nov 20 '16 at 07:39

score 0 · Answer 1 · answered Nov 20 '16 at 07:43

0

The algorithm you describe is not binary search.

To correctly implement binary search, keep track of two values, min_threshold and max_threshold. Initially set these to the minimum and maximum possible threshold values. Then at each step:

threshold = (min_threshold + max_threshold)/2
if GCC(threshold) > 60%:  # threshold is too low, update minimum
    min_threshold = threshold
else if GCC(threshold) < 50%:  # threshold is too high, update maximum
    max_threshold = threshold

answered Nov 20 '16 at 07:43

augurar

12,081
6
50
65

thnx i will try it now and write how it works at couple of hours .It is a 45Gb .dat file – valentinosael Nov 20 '16 at 07:46
`if GCC(threshold) > 60%:` is not valid python code BTW. – Jean-François Fabre Nov 20 '16 at 09:26
@Jean-FrançoisFabre Yes, this is pseudocode similar to what the OP used in his question. – augurar Nov 20 '16 at 09:30

Graph Giant Connected Component

1 Answers1