What is graph simulation? Modelling dynamic systems as graphs
"Graph simulation" means different things depending on who you ask. This article explains the meaning that matters if you want to model and simulate a real system: a graph simulation is a simulation whose model is a graph. Components are nodes that carry their own state, relationships are edges that describe how one component's state affects another's, and the simulation advances that whole structure through time.

Three meanings of the term
If you search for the phrase, you will find three unrelated ideas sharing a name.
- Graph simulation as pattern matching. In database and graph-algorithm research, a simulation is a relation between a small query graph and a large data graph. It is a cheaper alternative to subgraph isomorphism for finding where a pattern occurs. Nothing evolves over time.
- Simulating graphs as data. Statistics packages generate random, scale-free or small-world graphs to test algorithms. The graph is the output, not the model.
- Simulating a system described by a graph. The graph is the model of something that changes: heat flowing between parts, goods moving through a supply chain, money moving between banks. This is the meaning used here, and the one Konjugate is built for.
How a graph simulation works
A dynamic system is a set of things whose conditions change over time because they influence each other. A graph is a natural way to write that down.

- Nodes hold state. A hot plate has a temperature. A warehouse has a stock level. Each node owns its variables and nothing else.
- Edges hold relationships. An edge says how the state of one node pulls on another, usually as an equation. Heat moves from the plate to the air at a rate that depends on both temperatures.
- Time steps the whole graph. The engine evaluates every relationship, updates every node and repeats. The behaviour of the whole system emerges from local rules.
Because the structure is explicit, you can see which parts depend on which. You can swap one relationship without rewriting the model, and you can ask questions such as "what happens if this link fails?" by editing one edge.
Why model this way
Traditional simulation often hides structure inside one large block of equations or code. A graph keeps the structure visible and makes the model composable: subsystems can be built separately and connected afterwards. It also fits naturally with analysis that works on graphs, such as recovering the causal links in a system from measured data.
Graph simulation in Konjugate
Konjugate is open-source graph simulation software. You build the graph visually, write relationships as equations or in C++ or Python, and run it on a native C++ engine. A few places to go next:
- Build your first model in "Getting started with Konjugate", or try it in your browser without installing anything.
- Learn the canvas in "The graph canvas".
- Run a model and read the output in "Running a simulation" and "Exploring results".
- Learn how to organise a large graph in "Subsystems and edge groups".
- Recover a graph from data in "Recovering structure from data".
The source is on GitHub under the Mozilla Public License 2.0.
Follow Konjugate on LinkedIn for updates, or subscribe via RSS.
Comments
Post a Comment