unforum

[SpghEddy]

[Samari]

I'm of the belief that this is an exercise in graph theory. the list we were given is a set of edges and that GPS points in the same city are in fact the same vertex.

for the purpose of this, we assume that the website is being hosted on Aunt M's computer somewhere in SF because that is what Dana has told us. (I understand that the hosting is by rack space but it's only because they couldn't set up a server on adsl that could handle the beating we've given that box, so suspend your disbelief.) connect every edge that has a vertex of SF. then connect the edges who have a vertex with an edge already drawn. repeat until you can't draw any more edges. the result could show several interesting possible results. I think that we might be able to at least disregaurd any edges that aren't on the resulting graph.

if someone with access to good graphing software ::coughxnbomcough:: could do this out it would be great, otherwise i'll be doing it by hand which could take a long time.

[John Incognito]

John Incognito

[kalamazoo]

I point out that not all points are part of a single, connected graph.

1) For a connected graph, every node must have an edge that indirectly connects it to every other node.

2) This means that every node must satisfy one of the following conditions:
a) The node has more than one edge.
b) The node has one edge, but that edge must connect to a node with more than one edge.

It is easy to point out a pair of nodes that fail both conditions: Pittsburgh and Philadelphia are connected to each other and nothing else.

Therefore, not all points are part of a single, connected graph.

That said, an amazing number of points ARE part of a single, connected graph....

Cities with multiple edges:

* Hyannis, MA (2 edges)
* New York, NY (7 edges)
* Baltimore, MD (3 edges)
* Washington, DC (7 outbound edges; one internal node pair)
* Arlington, VA (2 edges)
* Alexandria, VA (2 edges)
* Woodbridge, VA (2 edges)
* Orange Beach, AL (2 edges)
* Gulf Shores, AL (2 edges)
* Kissimmee, FL (4 edges)
* Lafayette, LA (6 edges; one internal node pair)
* Indianapolis, IN (2 edges)
* St. Louis, MO (5 edges)
* Chicago, IL (1 edge; two internal node pairs)
* Sedalia, MO (7 edges)
* Madison, WI (2 edges)
* Boulder, CO (6 edges)
* Longmont, CO (2 edges)
* Salt Lake City, UT (7 edges)
* Phoenix, AZ (2 edges)
* Boise, ID (2 edges)
* Coeur d'Alene, ID (2 edges)
* Bellingham, WA (3 edges)
* Seattle, WA (9 edges)
* Kent, WA (3 edges)
* Portland, OR (8 edges)
* San Franscisco, CA (4 edges)
* Berkeley, CA (2 edges; one internal node pair)
* Palo Alto, CA (2 edges)
* Burbank, CA (3 edges)
* Los Angeles, CA (4 edges)
* San Diego, CA (4 edges)

Note that this list does not count points in suburbs... quite possibly there are more interconnections to be made there; but without an idea of what the PMs consider "close enough" I can't connect them.

Even so, the amount of connectivity is astounding: among the cities named, only three (Baltimore, Woodbridge, and Coeur d'Alene) are not part of a single, connected graph. Said graph involves 78 of the 110 edges.

By my classification:

* Edges unconnected to network, remains within one city: 2
* Edges unconnected to network, stretches between two cities: 23
* Edges connected, but not to main network: 7
* Edges connected to main network, remains within one city: 5
* Edges connected to main network, stretches between two cities: 73

Clearly the PMs/Melissa/SPDR are clustering these edges in about twenty cities for a reason. What that purpose is, is unclear. It's not relative to city importance: Sedalia, MO? It's in the middle of nowhere. Gulf Shores, AL? A seashore resort, and not a very big one. Boulder gets seven edges while Denver gets one. And there is a **distinct** bias in favor of Western cities (probably due to greater PM familiarity OOG, or proximity to Melissa's Castle in San Fransisco IG).

My thoughts? "Axons go hot" in relation to an AI would put one immediately in mind of a neural net. Could this be a neural net that Melissa is building?

[John Incognito]

Perhaps a city or town with only two nodes and a straight line between them is meant to reprensent the letter I or the number 1. Of course that's just assuming we really are playing connect the dots.

Extra thought on the matter: Maybe the cords/nodes/whatever spell out a message if you mark them on a map of the US.

Just brainstorming.

[WhiteGulls]

Okay, so I havn't seen anyone post on this, and this is just wild spec. So, the 21 links came from the same page as the coords, right? Well, my theory might work, even if they don't. Anyone tried overlaying addresses from WHOIS info with the addresses from the GPS coords? It might bring down the number of coords, or reveal something. I'm just a bit too busy to try it, and will be for a few days, so I figured I'd post it here, and see if anyone wants to give it a shot. If no one does, I'll probably be able to get started on doing that sometime around... wensday? And that's the soonest I could.

If this is trout, and someone else has already tried this sorry. Also, I had no idea where to post this, so I guessed that thsi thread might be the right place. If it isn't, could someone point me in the right direction? Thanks.

[Theq629]

Graph

[will_bushman]

Since this is in a large part an AI game, and has some familiar AI terminology in this puzzle, axons from neural nets, why not treat it as an old AI problem, the traveling salesman problem. Essentially try to determine the shortest path which goes through every point, but doesn't go through any point twice. connecting the dots in that order may provide us with an interesting picture. If someone has an algorithm for this sitting around from an AI class, that would be awesome.

[sherpa]

[strider]

Another lateral thought

[Clipped_Wings]

hey, finished my picture for GPS points<->SF, drawing lots of lines in photoshop is BORING.

Anyway, I noticed 3 things...

1. Points in important areas tend to be clumped

2. There are many instances where points lie on the same line

3. There are a few where that ISNT the case.

I now have a new idea...

The spider went out building roads and the coordinates are where he "popped his head up to see where he was".

Most roads keep going to the edge of america, but some stop around the middle, I think these are the ones that the spider found something of interest on.


(I'm going to post the results of both is and my next escapade both here and in relative, because I think it's relevant to both.)

[Clipped_Wings]

I think I've narrowed down how many points are useful.

It came to me when I was thinking of that old phrase "it's always in the last place you look".

If the lines are roads then most lead to the edge of america, those that stop short would imply something was found there.

I think these points are the ones worth investigating.

[b167er]

Neural Net Theory

[sherpa]

I implemented a Tic-Tac-Toe learning neural net last year, and I'm pretty sure it had less than 220 nodes - it only had two intermediate layers, too, IIRC. The 'Go' learning one I wrote was sizeably bigger, though, and Melissa's way more complex than even that, obviously.

At any rate; I don't think these axons are the whole neural network. I think they're an extension of it, in some way.

[Nola]

Could the data pairs represent polar coordinates (Radii, angle) instead of normal coordinates (x,y)? Has anyone tried graphing them that way?