Hello!
I have the following heterogenous graph -
One of the vertices represent users and weighted edges between these user vertices represent the confidence we have in those users being the same.
Now here's my question:
Say I am starting from user A and finding all linked user nodes to this user.
Say the traversal looks like A -> B -> C -> D.
And the weights look like thus:
[A, B] : 0.6
[B, C] : 0.5
[C, D] : 0.3
If I am trying to find all nodes linked to A with a confidence of at least 0.5, would we multiple the confidence scores or just take the min value?
So, option A if we consider multiplication of scores
A -> B : 0.6
A -> C : 0.3
A -> D : 0.09
In this case, given a threshold of 0.5, we only consider B as a close match of the user.
Option B: If we consider a min score,
A -> B: 0.6
A -> C : Min (0.6, 0.5) = 0.5
A -> D: Min (0.6, 0.5, 0.3) = 0.3
In this case, given a threshold of 0.5, we'd consider B and C to be a close match of the user A.
Which one would make more sense?
Thank you!
- Badges
- Science topic