Can diameter be zero?

A graph with one node has diameter zero under the usual convention.

Can weighted diameter be non-integer?

Yes, when shortest-path distances use non-integer weights.

Is diameter the same as radius?

No. Diameter takes the maximum eccentricity; radius takes the minimum.

AI Glossary / Graphs and symbolic AI / Advanced

Diameter

Definition

In graph theory, the diameter is the greatest shortest-path distance between nodes in a graph. For a connected, undirected graph with unweighted edges, it is the largest number of edges needed to connect any pair when each pair uses a shortest route. Weighted and directed graphs require explicit distance and reachability conventions.

What it means

Diameter describes the most distant pair under the graph's chosen definition of distance. First find the shortest route for each pair; then select the largest resulting distance. It does not mean the longest route that someone could deliberately take through the network.
This distinction matters in AI systems that represent relationships, dependencies, or movement as graphs. A small diameter can indicate that nodes are separated by few links, but it does not establish that information travels reliably, people cooperate, or a process finishes quickly. Those conclusions require additional evidence about what the links represent.

How it works

A node's eccentricity is its greatest shortest-path distance to other nodes. The graph's diameter is the maximum eccentricity. Its radius, by comparison, is the minimum eccentricity. These quantities summarize different aspects of the same distance structure.
For unweighted graphs, shortest paths count edges. For weighted graphs, distances commonly add edge weights, which must represent a meaningful cost or length. A connection-strength score should not automatically be treated as distance: stronger connections may require an appropriate transformation before shortest-path analysis.
Disconnected graphs contain pairs with no connecting path. Depending on the mathematical convention or software, their whole-graph diameter can be infinite, undefined, or reported as an error. Measuring only the largest connected component is a different calculation and should be labeled accordingly. Direction introduces a similar issue when one node cannot reach another.

Three practical examples

Knowledge navigation: A team measures shortest-link distances in a connected concept graph. Diameter identifies the greatest separation within that representation. It cannot show whether intermediate links make a useful explanation or whether the graph's underlying facts are correct.
Workflow dependencies: Analysts study a directed graph of allowed transitions. They must decide whether reverse travel is meaningful and how unreachable pairs are handled. Treating the graph as undirected answers a different question from following the actual transition rules.
Communication planning: An organization models possible relays between connected devices. An unweighted diameter describes the greatest minimum hop count. Actual delivery time also depends on transmission delays, congestion, retries, and processing, so hop count alone is not a latency guarantee.

Think of it this way

Imagine asking every pair of stations for its shortest journey, then selecting the longest answer. That answer is the diameter under the chosen journey measure. It is not the longest sightseeing trip, and a station disconnected from the network cannot simply be assigned an ordinary journey length.

Worked example

Consider five nodes joined in an undirected path: A-B-C-D-E. Every edge has length one. The shortest route from A to E uses four edges, and no pair is farther apart. The graph's diameter is therefore 4.
The eccentricities of A, B, C, D, and E are 4, 3, 2, 3, and 4. Their maximum is 4, confirming the diameter. Their minimum is 2, giving the radius; C is a central node under this distance definition.
Now add an edge directly connecting A and E. The graph becomes a five-node cycle. Every node can reach every other node in at most two edges, so the diameter falls to 2. A simple path containing four edges still exists, demonstrating why longest simple path and diameter are different quantities.
For a separate weighted example, take a triangle with edge lengths A-B = 2, B-C = 3, and A-C = 10. The shortest A-to-C route goes through B and has length 5. The three pairwise distances are 2, 3, and 5, giving diameter 5, not 10.
These results follow from the stated graphs and weights. Adding an isolated node F to either connected graph creates unreachable pairs; reporting the previous diameter would then require explicitly restricting the measurement to the original component.

Illustrative case study

Fictional Cedar Knowledge maps links between its internal guidance pages. Its first report presents a diameter for the largest connected component but calls it the diameter of the whole knowledge base. Reviewers discover several isolated groups of pages that the calculation excluded.
The team revises the report to show component coverage alongside each component's distance summary. It also checks whether links are usable in both directions before adopting an undirected representation. Proposed links are evaluated for relevance, not just their ability to reduce the numerical diameter.
The revised analysis supports navigation improvements while preserving the distinction between graph structure and reader experience. This is an illustrative scenario, not a measured organizational outcome.

Common misconceptions and mistakes

“Diameter is the longest path.” It is the largest shortest-path distance.
“The heaviest edge determines weighted diameter.” A cheaper alternate route may bypass that edge.
“A small diameter means every task is fast.” Distance represents only the chosen graph measure.
“Disconnected nodes can be ignored without explanation.” Excluding them changes the population being measured.
“Directed and undirected diameter are interchangeable.” Reachability and distances can change when direction is removed.

Business use

Define nodes, links, direction, weights, and exclusions before comparing diameters. Report the component covered and the measurement units. If the graph changes over time, check whether a reported improvement comes from better connections or merely from removing difficult-to-reach nodes.
Use diameter alongside coverage and typical-distance measures. Because it reflects an extreme pair, one peripheral node can affect it substantially. Whether that node deserves attention depends on its role in the real service, not just its position in the graph.

Check your understanding

Why is the five-node path's diameter 4?

Answer: Its most distant endpoints need four edges along their shortest route.

Why is the weighted triangle's diameter not 10?

Answer: The alternate A-B-C route costs only 5.

What must accompany a component-only result?

Answer: A clear statement of which component and nodes it covers.

Related terms

Shortest Paths
Edge weight
Shortest path
eccentricity
radius
graph center
connected component
reachability
path length

Learn next

Draw a connected graph and calculate each node's eccentricity. Add a useful connection, recalculate diameter and radius, and explain which shortest routes changed and which practical conclusions remain unsupported.

Sources

NetworkX, diameter reference: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.distance_measures.diameter.html NetworkX, eccentricity reference: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.distance_measures.eccentricity.html

Previous: Dialogue act Next: Dichotomic search