Space requirements for distance estimates in sparse graphs
Researchers found new lower bounds for pivot-based approximate distance oracles.
Researchers studied approximate distance oracles (ADO). An ADO is a compact data structure that answers distance queries within a multiplicative stretch .
A pivot-based -hop ADO gives every vertex pivots. It returns an estimate of the length of a walk .
In this walk, is a pivot of and is a pivot of . The distance is a stored distance.
Under the Erdős girth conjecture for graphs with constant degree, every pivot-based -hop ADO with stretch requires space. This result holds for every real .
The result also extends to pivot-based ADOs with any constant number of hops. These require space.
Using high-girth graphs of Double Agent [doubleAI'26], unconditional versions of these ADO lower bounds are obtained. The exponent above is multiplied by .
Key facts
- Pivot-based -hop ADOs require space under the Erdős girth conjecture.
- The lower bound for pivot-based ADOs with hops is .
- Unconditional lower bounds using Double Agent [doubleAI'26] have an exponent multiplied by .