Algorithm extends 2-approximation to nearby vertices in massive networks

Overview

Updated 2 hours ago

For decades, a 1996 algorithm could estimate distances across a huge network within twice the true value, but only when the points were far apart. For nearby points, the guarantee could fail.

Manoj Gupta of IIT Gandhinagar presented a refinement at FOCS 2025 that samples network connections at multiple scales. It extends the 2-approximation promise to close vertex pairs while keeping the earlier runtime.

Why it matters

Reliable, fast distance estimates matter for maps, communications and biology, where exact all-pairs calculations can overwhelm computers.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

2×

Approximation guarantee

Estimates fall between the true distance and twice that distance.

1996

Year of the DHZ algorithm

Dor, Halperin and Zwick published the method Gupta refines.

FOCS 2025

Presentation venue

Gupta presented the paper at the 66th Annual Symposium on Foundations of Computer Science.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Manoj Gupta

Associate Professor, IIT Gandhinagar

Presented the improved 2-approximation at FOCS 2025

Organizations Involved

Indian Institute of Technology Gandhinagar

University

Gupta's institution

The university where Manoj Gupta works and where the research was based.

IEEE Annual Symposium on Foundations of Computer Science

Academic conference

Venue for Gupta's paper

A peer-reviewed conference where Gupta presented the improved algorithm.

Timeline

1996 October 2026

3 events Latest: 2 days ago

October 2026
ScienceDaily reports the algorithm breakthrough
Latest Publication

The report describes the algorithm as theoretical and not yet a commercial system.
2 days ago
2025
Gupta presents close-pair algorithm at FOCS 2025
Research

Multi-scale sampling extends the 2-approximation guarantee to nearby vertices without added runtime.
2025
1996
Dor, Halperin and Zwick publish 2-approximation
Research

Their method estimates long distances within twice the true value in nearly optimal time.
1996

Scenarios

Follow-up paper lowers the 2-approximation factor for close pairs

Possible Resolves by End of 2028

Discussed by: ScienceDaily and IIT Gandhinagar coverage

A follow-up paper could push the guarantee below twice the true distance for close vertex pairs. The trigger is a new sampling or rounding step that preserves near-optimal runtime.

Gupta's sampling method reaches a mainstream graph library

Unlikely Resolves by End of 2028

Discussed by: Open-source graph software maintainers

A major graph library could add an approximate all-pairs routine based on multi-scale sampling. Adoption would require an implementation that beats existing methods on large sparse graphs.

Historical Context

2 moments from history that rhyme with this story — and how they unfolded.

1996

Dor-Halperin-Zwick 2-approximation (1996)

Dor, Halperin and Zwick showed that a small set of sampled vertices can estimate distances within twice the true value in nearly optimal time. The method worked well when the points being compared were far apart.

Then

The method became a standard approximation for distant pairs.

Now

Its guarantee did not reliably cover close pairs, a limit that lasted until Gupta's refinement.

Why this matters now

Gupta's paper directly extends this line of work to nearby vertices.

1956-1959

Dijkstra's shortest-path algorithm (1956-1959)

Edsger Dijkstra developed a method to find the shortest path from one source vertex to every other vertex in a graph. It became a foundation of routing and network software.

Then

It made exact single-source shortest paths routine.

Now

Running it for every source is costly on dense networks, which is why all-pairs approximation matters.

Why this matters now

The new algorithm attacks the same all-pairs problem with approximation rather than exactness.