A General Reduction from Near-Additive Emulators to Near-Exact Hopsets

Abstract: Graph emulators and hopsets are two fundamental concepts for distance approximation. When the multiplicative stretch is $1+ε$ for arbitrarily small $ε>0$, these structures are known as near-additive emulators and near-exact hopsets, respectively. Prior work showed that there is a remarkable similarity between the constructions and guarantees of these two objects. In their survey on this topic, Elkin and Neiman [Bull. EATCS 130, 2020] explicitly asked whether one can obtain a general reduction between near-additive emulators and near-exact hopsets. Following that, Kogan and Parter [FOCS, 2022] provided a general reduction from hopsets to emulators and spanners. In this paper, we address the reverse direction and show that any construction for a near-additive emulator for undirected unweighted graphs can be leveraged as a black box to construct a hopset for an undirected weighted graph with comparable size, stretch, and a hopbound comparable to the emulator's additive stretch. Specifically, we show that any algorithm that constructs a $(1+ε',β)$-emulator, with $0 \le ε' \le 1$ and $β\ge 1$, of size $S_{\mathcal{A}}(n, ε',β)$, can be used to obtain a $$(1+ε, O(\frac{β^{2}{ε^2}} \ln(\frac{n}ε)))$$-hopset of size $$O((S_{\mathcal{A}}(n+m\fracβ{ε^2}, \fracε{294},β) \frac{1}ε + n)\ln(\frac{n}ε))$$, for any $0 < ε\le 1$. Therefore, our reduction answers the question of Elkin and Neiman [Bull. EATCS 130, 2020] for sparse graphs and further advances the understanding of the formal connection between these two structures. Designing a reduction resulting in a hopset size that does not depend on $m$ remains an intriguing open question.

• The paper provides a novel reduction from near-additive emulators to near-exact hopsets, preserving near-optimal stretch with only a modest increase in parameters.
• It employs multi-scale graph subdivisions and systematic projection techniques to ensure accurate distance approximations while controlling the number of hops.
• The construction yields explicit quantitative bounds linking emulator and hopset sizes, offering practical insights for sparse undirected weighted graphs.

General Reduction from Near-Additive Emulators to Near-Exact Hopsets

Introduction and Context

The paper "A General Reduction from Near-Additive Emulators to Near-Exact Hopsets" (2607.07190) systematically addresses the formal connection between two pivotal graph-theoretic structures used for distance approximation in algorithmic graph theory: near-additive emulators and near-exact hopsets. The motivation arises from both significant theoretical interest (with prior constructions and partial connections known) and explicit open questions in the literature, particularly the question posed by Elkin and Neiman concerning whether a general reduction from near-additive spanners or emulators to near-exact hopsets exists for sparse graphs. The paper provides a constructive and general affirmative answer for emulators, which directly implies similar results for spanners.

Core Definitions

• $(\alpha,\beta)$-emulator: A graph on the same vertex set as $G$, possibly with non-subset edge set, which preserves all pairwise distances up to multiplicative stretch $\alpha$ and additive term $\beta$.
• $(\alpha,\beta)$-hopset: A collection of weighted edges added to $G$ so all pairs admit approximate shortest paths with at most $\beta$ hops and stretch at most $\alpha$.
• Near-additive: $\alpha = 1 + \epsilon$ for arbitrarily small $\epsilon > 0$; near-exact is the analogous notion for hopsets.
• The strong relationships and partial constructions formalized in prior work (e.g., [Elkin & Neiman 2019], [Huang & Pettie 2019], [Neiman & Shabat 2022]) set the background.

Main Results and Theoretical Contributions

The General Reduction

The principal result is a generic black-box reduction: any algorithm that constructs a near-additive emulator for undirected unweighted graphs yields a construction of a hopset for undirected weighted graphs with quantitatively related parameters. Specifically, if there exists an algorithm producing a $G$0-emulator of size $G$1, then, for any weighted undirected graph $G$2, the construction in the paper yields a $G$3-hopset of size $G$4, for any $G$5.

This resolves the previously open direction, complementing the more direct reduction from hopsets to emulators given in Kogan and Parter [FOCS 2022].

Quantitative Guarantees

• Hopbound: Quadratic dependence on emulator additive stretch $G$6 and $G$7 (i.e., $G$8).
• Size: $G$9, where $\alpha$0 is the size of the underlying emulator.
• Stretch is nearly preserved; only a $\alpha$1 multiplicative inflation is incurred.

Although the bound on hopset size includes an $\alpha$2 (edge-count) dependency, the authors clarify that this does not asymptotically impact most algorithmic applications, since runtimes already scale linearly with $\alpha$3.

Construction Techniques

The reduction's construction is organized by distance scales: for each scale, edges are subdivided so that all have weights at most $\alpha$4, and emulator constructions are executed on the resulting unweighted 'scale graphs'. For every hopset edge, the construction ensures that a path of the desired stretch and hopbound can be simulated, even when shortcuts may traverse across scales. A careful projection step maps edges incident to subdivision vertices back to the original vertex set in such a way that the approximation guarantees are preserved. The projection is handled via a systematic offset-based computation, ensuring that no path underestimates true distances.

The construction draws on the iterative shortcutting strategy similarly employed in parallel/dynamic/directed hopset constructions and prior emulation-to-hopset reductions.

Numerical and Structural Claims

The main numerical claims established by the reduction are:

• Hopset size tracks the emulator size up to a moderate blowup: if an emulator is $\alpha$5 edges, the hopset is only polylog larger.
• Stretch is essentially the emulator's multiplicative stretch, plus $\alpha$6.
• Hopbound is quadratic in the emulator's additive stretch and inversely in $\alpha$7.
• No underestimation: The construction ensures all hopset edges retain or overestimate graph distances.

A potentially limiting feature is that the hopset size remains $\alpha$8-dependent; removing this dependency is forwarded as a significant open question.

Implications and Future Directions

Practical implications are immediate: given access to any near-additive emulator construction, one can now systematically derive near-exact hopsets of competitive size and hopbound, with only modest loss in approximation quality or sparsity. This abstracts and unifies previous ad-hoc constructions observed in the literature, and enables the translation of future improvements in emulator constructions directly to hopset constructions in undirected weighted settings. The result explicitly resolves the open problem in the direction 'emulator $\alpha$9 hopset' for sparse undirected graphs.

Theoretically, this reduction closes a significant gap in the formal understanding of the relationship between distance-approximating graph structures, and the tools developed for the projection and analysis of paths with respect to subdivisions and multi-scale emulation may have utility in related settings (dynamic, distributed, and parallel algorithms).

Future work aimed at eliminating the $\beta$0-dependence in hopset size, tightening the hopbound dependence to match the best known lower bounds, and possibly generalizing analogous reductions for directed graphs or specialized spanner/emulator classes, would have substantial impact.

Conclusion

This work delivers a formal and quantitative reduction from near-additive emulators to near-exact hopsets, providing a definitive and general construction for undirected weighted graphs. The reduction precisely characterizes how emulator parameters propagate to hopset guarantees, establishing both theoretical clarity and practical utility. Open directions include optimizing the dependencies (notably, eliminating terms that scale with edge count $\beta$1) and extending the paradigm to broader graph classes and algorithmic models (2607.07190).