My grad-student career took me into Monte Carlo methods and viscosity-free fluid flow. It’s a respectable path. But I could have ended up in graph theory; I got a couple courses in it in grad school and loved it. I just could not find a problem I could work on that was both solvable and interesting. But hints of that alternative path for me turn up now and then, such as in this piece from 2018.
I am surprised to have had no suggestions for an ‘O’ letter. I’m glad to take a free choice, certainly. It let me get at one of those fields I didn’t specialize in, but could easily have. And let me mention that while I’m still taking suggestions for the letters P through T, each other letter has gotten at least one nomination. I can be swayed by a neat term, though, so if you’ve thought of something hard to resist, try me. And later this month I’ll open up the letters U through Z. Might want to start thinking right away about what X, Y, and Z could be.
This is another term from graph theory, one of the great mathematical subjects for doodlers. A graph, here, is made of two sets of things. One is a bunch of fixed points, called ‘vertices’. The other is a bunch of curves, called ‘edges’. Every edge starts at one vertex and ends at one vertex. We don’t require that every vertex have an edge grow from it.
Already you can see why this is a fun subject. It models some stuff really well. Like, anything where you have a bunch of sources of stuff, that come together and spread out again? Chances are there’s a graph that describes this. There’s a compelling all-purpose interpretation. Have vertices represent the spots where something accumulates, or rests, or changes, or whatever. Have edges represent the paths along which something can move. This covers so much.
The next step is a “directed graph”. This comes from making the edges different. If we don’t say otherwise we suppose that stuff can move along an edge in either direction. But suppose otherwise. Suppose there are some edges that can be used in only one direction. This makes a “directed edge”. It’s easy to see in graph theory networks of stuff like city streets. Once you ponder that, one-way streets follow close behind. If every edge in a graph is directed, then you have a directed graph. Moving from a regular old undirected graph to a directed graph changes everything you’d learned about graph theory. Mostly it makes things harder. But you get some good things in trade. We become able to model sources, for example. This is where whatever might move comes from. Also sinks, which is where whatever might move disappears from our consideration.
You might fear that by switching to a directed graph there’s no way to have a two-way connection between a pair of vertices. Or that if there is you have to go through some third vertex. I understand your fear, and wish to reassure you. We can get a two-way connection even in a directed graph: just have the same two vertices be connected by two edges. One goes one way, one goes the other. I hope you feel some comfort.
What if we don’t have that, though? What if the directed graph doesn’t have any vertices with a pair of opposite-directed edges? And that, then, is an oriented graph. We get the orientation from looking at pairs of vertices. Each pair either has no edge connecting them, or has a single directed edge between them.
There’s a lot of potential oriented graphs. If you have three vertices, for example, there’s seven oriented graphs to make of that. You’re allowed to have a vertex not connected to any others. You’re also allowed to have the vertices grouped into a couple of subsets, and connect only to other vertices in their own subset. This is part of why four vertices can give you 42 different oriented graphs. Five vertices can give you 582 different oriented graphs. You can insist on a connected oriented graph.
A connected graph is what you guess. It’s a graph where there’s no vertices off on their own, unconnected to anything. There’s no subsets of vertices connected only to each other. This doesn’t mean you can always get from any one vertex to any other vertex. The directions might not allow you to that. But if you’re willing to break the laws, and ignore the directions of these edges, you could then get from any vertex to any other vertex. Limiting yourself to connected graphs reduces the number of oriented graphs you can get. But not by as much as you might guess, at least not to start. There’s only one connected oriented graph for two vertices, instead of two. Three vertices have five connected oriented graphs, rather than seven. Four vertices have 34, rather than 42. Five vertices, 535 rather than 582. The total number of lost graphs grows, of course. The percentage of lost graphs dwindles, though.
There’s something more. What if there are no unconnected vertices? That is, every pair of vertices has an edge? If every pair of vertices in a graph has a direct connection we call that a “complete” graph. This is true whether the graph is directed or not. If you do have a complete oriented graph — every pair of vertices has a direct connection, and only the one direction — then that’s a “tournament”. If that seems like a whimsical name, consider one interpretation of it. Imagine a sports tournament in which every team played every other team once. And that there’s no ties. Each vertex represents one team. Each edge is the match played by the two teams. The direction is, let’s say, from the losing team to the winning team. (It’s as good if the direction is from the winning team to the losing team.) Then you have a complete, oriented, directed graph. And it represents your tournament.
And that delights me. A mathematician like me might talk a good game about building models. How one can represent things with mathematical constructs. Here, it’s done. You can make little dots, for vertices, and curved lines with arrows, for edges. And draw a picture that shows how a round-robin tournament works. It can be that direct.