Certificate-Sensitive Subset Sum: Realizing Instance Complexity

2025-07-21

AI Summary

This paper addresses the NP-complete Subset-Sum problem and introduces the first **certificate-sensitive deterministic algorithm**, whose runtime depends on the **structural complexity certificate size** of the input instance—departing from traditional worst-case analysis. The method combines deterministic enumeration with structured pruning, leveraging algebraic properties of the subset-sum set Σ(S). It solves Subset-Sum in O(U·n²) time and O(U·n) space, where U is an upper bound on the target value. Compared to classic meet-in-the-middle search (O*(2^{n/2})), our algorithm is strictly faster in the worst case (O*(2^{n/2−ε})) and achieves **structure-adaptive efficiency**: it significantly accelerates on simple instances while retaining theoretical guarantees for hard ones. This work establishes the first input-dependent deterministic adaptive framework for Subset-Sum, marking a conceptual shift toward certificate-driven algorithm design for NP-hard problems.

Technology Category

• Search and Optimization: Combinatorial Optimization
• Planning, Routing, and Scheduling: Deterministic Planning
• Constraint Satisfaction and Optimization: Satisfiability

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

We present, to our knowledge, the first deterministic, certificate-sensitive algorithm for a canonical NP-complete problem whose runtime provably adapts to the structure of each input. For a Subset-Sum instance $(S, t)$, let $Σ(S)$ denote the set of distinct subset sums and define $U = |Σ(S)|$. This set serves as an information-theoretically minimal witness, the instance-complexity (IC) certificate. Our solver, IC-SubsetSum, enumerates every element of $Σ(S)$ in deterministic time $O(U cdot n^2)$ and space $O(U cdot n)$. A randomized variant achieves expected runtime $O(U cdot n)$. The algorithm's complexity is thus directly governed by the certificate size, and this structure-sensitive performance is paired with a guaranteed worst-case runtime of $O^*(2^{n/2 - varepsilon})$ for some constant $varepsilon > 0$, the first such result to strictly outperform classical methods on every instance. We revisit fine-grained reductions that rely on the classical $2^{n/2}$ hardness of SubsetSum and show that these arguments hold only for collision-free instances where $U$ is maximal. IC-SubsetSum reframes this barrier structurally and introduces a new paradigm for certificate-sensitive algorithms across NP-complete problems.

Problem

• Develops deterministic algorithm adapting to input structure
• Solves Subset-Sum with complexity tied to certificate size
• Outperforms classical methods on all instances

Innovation

• Deterministic algorithm adapts to input structure
• Enumerates subset sums in O(U*n^2) time
• Worst-case runtime strictly outperforms classical methods