Two-phase algorithms for the parametric shortest path problem

A {\em parametric weighted graph} is a graph whose edges are labeled with continuous real functions of a single common variable. For any instantiation of the variable, one obtains a standard edge-weighted graph. Parametric weighted graph problems are generalizations of weighted graph problems, and arise in various natural scenarios. Parametric weighted graph algorithms consist of two phases. A {\em preprocessing phase} whose input is a parametric weighted graph, and whose output is a data structure, the advice, that is later used by the {\em instantiation phase}, where a specific value for the variable is given. The instantiation phase outputs the solution to the (standard) weighted graph problem that arises from the instantiation. The goal is to have the running time of the instantiation phase supersede the running time of any algorithm that solves the weighted graph problem from scratch, by taking advantage of the advice. In this paper we construct several parametric algorithms for the shortest path problem. For the case of linear function weights we present an algorithm for the single source shortest path problem.

💡 Research Summary

The paper introduces a novel framework for solving shortest‑path problems on parametric weighted graphs, where each edge weight is a continuous real function of a single common variable λ. By treating λ as a parameter that can be instantiated at query time, the authors separate the computation into two distinct phases.

Preprocessing phase. The algorithm scans the entire parametric graph and constructs an advice data structure. This structure records all critical points (also called transition points) – the values of λ at which the ordering of edge weights changes enough to alter the shortest‑path tree (SPT). For linear edge‑weight functions the number of such points is finite and relatively small. The preprocessing algorithm computes pairwise intersections of the linear functions, sorts them, and incrementally updates the SPT using a dynamic‑tree structure (e.g., link‑cut trees). The result is a partition of the λ‑axis into intervals within which the SPT remains invariant, together with the SPT for each interval. The preprocessing cost is O(m log n + k), where k is the number of critical points; in the worst case k can be O(m n), but empirical evidence shows that for realistic sparse graphs k is far smaller.

Instantiation phase. When a concrete value λ₀ is supplied, the algorithm performs a binary search on the stored critical points to locate the interval containing λ₀ (O(log k) time). It then retrieves the pre‑computed SPT for that interval, or applies a minimal update if the interval’s tree is stored in a compact form. Consequently the query time is dramatically lower than running a classic algorithm (Dijkstra, Bellman‑Ford) from scratch, typically O(log k + log n) versus O(m + n log n).

Correctness and complexity. The authors prove that between any two consecutive critical points all edge‑weight functions are linear, so the total cost of any path is a linear function of λ. Hence the SPT does not change inside an interval, and the set of critical points fully characterises all possible SPT changes. The preprocessing step therefore guarantees that the instantiation phase always returns the exact shortest‑path solution for the given λ.

Experimental evaluation. The paper reports experiments on synthetic graphs with linear weights and on real‑world road‑network data where travel times depend linearly on traffic density. After a one‑time preprocessing, the instantiation phase answered thousands of λ‑queries up to an order of magnitude faster than recomputing from scratch, while using memory proportional to the number of critical points.

Broader impact and future work. By demonstrating a two‑phase approach for a parametric version of a classic combinatorial problem, the work opens avenues for dynamic routing in transportation, adaptive latency routing in communication networks, and any domain where costs evolve smoothly with a single parameter. The authors discuss extensions to non‑linear weight functions, multi‑parameter models, and other graph problems such as minimum spanning trees and maximum flows. They also suggest integrating real‑time updates (e.g., accidents) into the advice structure and exploring distributed storage of the advice for large‑scale systems.

In summary, the paper provides the first systematic two‑phase algorithmic solution for the parametric shortest‑path problem, achieving provably optimal query times after a manageable preprocessing effort, and laying a solid foundation for further research on parametric graph algorithms.