Adjacency matrix to a weighted directed graph

Hi guys, I'm trying to convert a given adjacency matrix to a visualised weighted directed graph in rhino and feel a little out of my depth. Does anyone know a way I can do this?

The final result should show something like this image to the right in rhino.

The methodology doesn't really matter, as long as it can be applied to any given adjacency matrix.

If anyone solves this Ill send them some sweets. Cheers!

Replies are closed for this discussion.

Replies to This Discussion

Permalink Reply by Tiara Jacinta on October 12, 2015 at 5:37pm

Forget all the rules and the output it flexible. The only thing that needs to be a constant is the location of the Judge. This point must be stable. The other locations and orientations of the participants are flexible.

In practise this could work with a series of string and magnets. The various forces depending on which direction they are faced would determine the type of force(either repellent or attraction).

Ill try and start small, make three points and attach forces of different magnitude to two of the points and direct them towards one point then get the forces to summarise.

We could build from here.

Permalink Reply by Tiara Jacinta on October 12, 2015 at 6:50pm

This is sort of what I mean but instead of using lines you would use vectors... and not rotate the object sideways.. yeah.

Attachments:

lines as vectors.gh, 16 KB

Permalink Reply by peter fotiadis on October 13, 2015 at 12:19am

Let's reload the Matrix (again):

1. Adj matrices are the "rules" (weighted graphs are the visual interpretation of Matrices, not having any real-life usage AT ALL).

2. Cluster analysis (HAC or other) is the "grouping" of rules.

3. Dendrograms are the visual interpretation of 2 - or some sort of GUI for the 4.

4. In order to "translate" 3 into topology of some sort you need a totally different approach (than vector logic): you start (bottom-top) "combining" leafs (i.e. the most "connected" parameters) and for each climb in the tree (towards the root) you combine the "lower order" result (the most strongly connected sub-clusters) VS clusters with the equivalent order in the dendrogram's hierarchy. "Kinda" like recursion used in fractal type of algorithms (this is a thingy where a method calls itself: a bit freaky concept, he he).

5. 4 gives you no guaranties -of any kind - that the best solution can been found in a deterministic way of thinking. It can produce a range of possible solutions that are prone to evaluation/validation.

Steps 4-5 require code that is most unlikely that you can find anywhere. Step 3 is either doable from scratch or available in (paid - around 1-2K) libraries (NMath and the likes):

Why vector logic is out of question? Because in 99.9999% of cases we are talking about spaces (but in your oversimplified case: seats) that must comply with a lot of other restrictions as well.

You can attempt to approximate all that with vectors (hope always dies last) ... but is 1M miles away from the proper approach.

Moral: the most important thread in GH Forums (ever) ... is in a clear danger to get absorbed by the black hole.

Permalink Reply by peter fotiadis on October 13, 2015 at 12:30am

At that delicate point here's 2 abstract cases for you:

1. The K-Means clustering demo includes 3 challenges (for the brave). Master them and win the 1st prize (lot's of sardine cans, what else?)

Challenge 3 is the most important. Why you get (occasionally) this:

instead of that?

2. The HAC ... well is an abstract and very slow thingy, he he.

may the Force (the dark option) be with us all

Attachments:

clusterAnalysis_HAC_V1.gh, 129 KB
cLusterAnalysis_K-Means_V1A.gh, 140 KB

Permalink Reply by Tiara Jacinta on October 13, 2015 at 6:44pm

Working it out!

The cloud of number of points must be divisible by its clustering mechanism.

ie: 600 is not divisible by 7, so you get that really awkward result.

Permalink Reply by Tiara Jacinta on October 13, 2015 at 10:43pm

I'm having trouble visualising the result. Would the end result look something like this? If so, why can't we introduce force vector graph drawing?

It makes a whole lot more sense then using your method which doesn't have any source code.

Also the HAC script isn't showing anything. Do I need to do something to it?

Permalink Reply by peter fotiadis on October 13, 2015 at 11:57pm

It's about time for Plan Z: Skype

Permalink Reply by peter fotiadis on October 14, 2015 at 12:56am

Let's switch this talking off-line (use my mail just send). This topic it's rather too "focused" (but I still rate your thread as the most important ever in GH Forums, for Architects that is) and I guess that nobody else finds it interesting.