Faster Algorithms for Subset Sum using Filtered Representation Trees
Year: 2026
Type: preprint
Abstract: We study filtered representation algorithms for random planted subset sum on $n$ input numbers at density one. Within the binary representation-tree frameworks compared here, we conditionally improve the classical running time from $2^{0.2690n+o(n)}$ of Howgrave-Graham to $2^{0.2634n+o(n)}$. With quantum access to classical memory (QRACM), the running time improves from $2^{0.2356n+o(n)}$ to $2^{0.2100n+o(n)}$. With quantum access to quantum memory (QRAQM), it improves from $2^{0.2156n+o(n)}$ to $2^{0.1980n+o(n)}$. For each new construction, total space satisfies the same upper bound as running time. The constructions combine coordinate-dependent representation laws with type-class filtering and cost models adapted to their search engines. The quantum-memory construction uses one common size for all nonroot lists, so its remaining resource rates follow from short entropy formulas. We derive every source, index, candidate and reporting cost and verify the three finite profiles using directed interval calculations. The classical and quantum-search bounds require assumptions on recursive list distributions, survival and access. The quantum-memory bound further requires a coherent-update and hard-workspace interface, whose validity is not implied by an ordinary expected update count.
Authors: Adams, Samuel J., Iglesias, Luis
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235297879
