Skip to content

Latest commit

 

History

History

integer-factorization

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 

Conditional deterministic integer factorization

Assuming the cited Boolean tensor-contraction theorem, a uniform classical deterministic algorithm completely factors every positive integer with subpolynomial dependence on its numerical value. The new recovery constructions need at most one contraction and provide distinct precision/auxiliary-work tradeoffs. The assumption is visible in the theorem and every headline.

Complete proof · Exact dependency interface · Finite verification · Separate research branches

Objective and output

Return the complete list of prime factors and multiplicities of an arbitrary positive binary integer. Powers of two, prime powers, unbalanced semiprimes and inputs with many factors are included. The trivial input returns the empty factorization. At most one nontrivial contraction is needed.

Randomness

Conditional uniform deterministic classical bit algorithm. Exact transform-field construction, root recovery and factor extraction use no randomness. Correctness of the cited external tensor-contraction theorem is assumed; local finite verification does not establish it.

Time

With n=bit_length(N), for sufficiently large inputs,

Formula

The proof includes all preprocessing and complete-factor recovery. Bounded inputs are handled directly. This is subexponential in input bit length and remains superpolynomial; it is not a practical running-time estimate.

Three one-contraction choices retain this leading bound. The basic moment construction uses linear precision, logarithmic truncation and auxiliary work

Formula

The Sidon second-moment construction instead uses

Formula

The packed square-zero construction uses

Formula

with preprocessing and recovery costs

Formula

respectively. These options optimize different resources. The later constructions improve contraction count and coefficient precision, not the leading total-time exponent.

Space

Polynomial writable bits in the original input bit length, including preprocessing, the assumed contraction algorithm and complete recovery. All large gcd candidate families and trial-division candidates are streamed. No root list or local tensor truth table is expanded in the basic moment route. The packed route recovers only a polynomially bounded rough-residual root set.

Parameters

N>=1 is the numerical input and n=bit_length(N) its binary length. P is integer coefficient precision; ell is polynomial truncation length in (Z/2^P Z)[V]/(V^ell). The network has O(n) Boolean edges and polynomial succinct description length. Local transform-field arithmetic has polynomial bit cost and is part of the Boolean entry predicates.

Model and representation

The exact external theorem accepts bipartite Boolean multigraphs with ordered ports, at most linearly many edges, degree bounded by ceil(s^(9/10)), and uniform succinct entries in its literal truncated-integer ring, with precision and truncation at most a fixed constant times s. The construction meets this interface with s=Theta(n). The external theorem is assumed correct; it is not re-proved or implemented by this repository's finite verification programs. The full local reduction and extraction arguments are provided.

Lower-bound scope

No general factoring lower bound or optimality claim is made. The accompanying obstructions concern the specified retuning and interpolation templates. The asymptotic conditional result is significant relative to the cited general deterministic positive-exponent bounds. It remains asymptotically slower than the cited rigorous randomized subexponential algorithm. Publication-wide priority and practical cryptanalytic performance are not certified.

Review status

Ready for review. The complete conditional classical deterministic factoring reduction and one-contraction recovery family: the original linear-edge Fourier multiplication/prefix reduction; three-forward-transform unique-extension root verifier; signed digit-moment separation and saturated unitary gcd refinement; basic, Sidon square-root-precision and packed subpolynomial-precision resource choices; deterministic characteristic-two field construction, monotone regular-representation paths, literal-ring guarded output and Hankel/trace root recovery. The actual external theorem interface and the scoped template limitations were checked.

Dependencies and scope. The cited OpenAI October 25, 2026 succinct Boolean tensor-contraction theorem is an explicit accepted premise. This review checks its literal applicability and proves the local reduction conditional on it; it neither re-proves the external theorem nor validates its implementation. The result is deterministically subpolynomial in numerical input and subexponential, still superpolynomial, in binary input length. The exact theorem counts Boolean edges, has a fixed linear precision/truncation allowance and supports uniform succinct entry procedures. Internal transform-field and auxiliary square-zero algebra computations are compiled into Boolean predicates and literal-ring outputs; no broader coefficient-ring theorem is assumed. Original inputs need no balance or squarefreeness promise. The later increment improves contraction count and precision/recovery tradeoffs without improving the leading time exponent. Full-grid and full-packet qualifications limit the characteristic-two obstruction; no optimality or general factoring lower bound is claimed. The separate direct one-fifth-exponent candidate remains outside this selected theorem scope. Finite arithmetic programs check small instances and provenance; no complete fast contraction backend is shipped. Publication-wide priority, external peer review, practical cryptanalytic speed and proof-assistant verification are not certified.

The verification note identifies executable coverage. This status does not claim external peer review or proof-assistant verification.

Source

Canonical manuscripts and external dependencies. The two supplied source increments are mapped in the import record. All curated output is English. Source-derived mathematics is preserved; no intermediate checkpoints or stronger exponent improvement are inferred.

Attribution: icekylinx and GPT-6 Astra Pro.