Similarity between two vectors representing stellar graphs

Question

it's a bit algorithmic, and I'm not sure if it matches SO properly, let me know if it doesn't.

I want to implement some distance or similarity function between two marked asterisks. For example, I might have the following graphs view:

A = [3 -1 0; 4 1 0; 2 0 1; 2 0 -1] 

and

 B = [3 -0.8 1 4 1.4 0.4; 2 0.1 1.9] 

My idea (which is actually based mainly on the article I'm reading) is to do some kind of propagation of faith: let's say that A is an observed graph - each row is an adjacent center node, the first column is a label, the second column is the position of x, and the third is the position of y (the positions belong to the central node), B is the graph in the database, and I want to see how close or similar they are.

In the article, they calculate the probability that the observed graph A is given as the product of the probabilities for all nodes. How would you effectively implement this in Matlab?

I would be grateful for any thought or inspiration, I am not looking for a code to solve it, I just want to find out from you masters how you would approach this problem.

Thanks,

J

Answer 1

Having only a superficial look at this article, I think that you are looking for an output algorithm in Bayesian networks. There are, as you know, expensive and difficult to implement computing tools, so I recommend looking for ready-made solutions. Google returns many views for "Matlab graphical models"

- EDIT -

A website posted in one of the comments looks pretty detailed, but may be a bit dated (2005). I would like to recommend other software. I have been using MALLET for a while, and I find it quite normal. The authors are really well known in the NLP community. It turns out that there is an extension that allows an arbitrary graphical structure - GRMM . Written in Java, open source. I did not use it myself.