KEVOS
ArticlesServicesCase studiesAboutContact
ArticlesServicesCase studiesAboutContact
← ArticlesDetecting Algebraic and Linear Dependence with LLLEngineering · Engineering MathematicsLesson 754/884← PrevNext →
ArticlePublished 7 Aug 20262 min readBy Kevin Jogininteger relationalgebraic dependenceLLLPSLQ
On this page

Ask about this page

KEVOS AIDetecting Algebraic and Linear Dependence with LLL

KEVOS knowledge first · trusted web sources when needed

Lattices and LLL Reduction

Detecting Algebraic and Linear Dependence with LLL

Recovering exact integer relations from numerical approximations using LLL, and the precision requirements that make the method reliable.

Engineering / MathematicsLattices and LLL Reduction2 min readKV-MATH-0552

Given a real number known only to high precision, LLL can often recover an exact polynomial it satisfies. The technique is one of the most striking applications of lattice reduction and one of the easiest to misuse.

The construction

To find an integer relation among given real numbers, build a lattice whose vectors pair the candidate coefficients with a scaled approximation of the corresponding combination. A short vector corresponds to a small combination — that is, to a relation.

Rows of [ I ; round(c * x_i) ] for a large scale cA short vector has small coefficients and a nearly vanishing combination.

Integer relation detection

  1. Compute to high precisionObtain the values to substantially more digits than the expected coefficient size.
  2. Build the latticeIdentity block alongside scaled values.
  3. ReduceApply LLL.
  4. Read the candidateThe first vector gives candidate coefficients.
  5. Verify exactlyConfirm the relation by an independent exact computation. This step is mandatory.

Finding a minimal polynomial

Applying this to the powers of an algebraic number recovers its minimal polynomial: a relation among the powers is precisely a polynomial the number satisfies.

Key point

The degree must be guessed. Running the construction for increasing degrees and stopping at the first verified result is standard, but a relation found at too high a degree may be a multiple of the true minimal polynomial rather than the minimal polynomial itself.

Precision

Caution

Insufficient precision produces a spurious relation that satisfies the approximation but not the exact values. The output looks entirely plausible. Precision must exceed roughly the degree times the expected coefficient size in digits, and the verification step is not optional under any circumstances.

Diagnosing integer relation detection
SymptomDiagnosis
Relation fails exact verificationInsufficient precision
Coefficients as large as the scale factorNo genuine relation exists at this degree
Different relations at different precisionsSpurious; increase precision
Relation is a multiple of a smaller oneDegree guess was too high

Applications

The technique recovers minimal polynomials of algebraic numbers given numerically, identifies class invariants in complex multiplication, and supports polynomial reduction by recognising small defining polynomials.

Note

PSLQ is an alternative algorithm for the same problem with different numerical properties. Both are widely used; LLL is more general, PSLQ is often more stable for this specific task.

Source. Henri Cohen, A Course in Computational Algebraic Number Theory, Springer GTM 138 — 2.7.2. Structural reference unverified: the source file was not available during authoring; chapter and section numbers are taken from the published edition and have not been checked against a physical copy.

Related pages

  • The Polynomial Reduction Algorithm
  • Complex Multiplication and Class Numbers
  • Integer Kernel and Image via LLL
  • Finding Short Vectors in Lattices

Continue learning

Integer Kernel and Image via LLLArticle · Engineering MathematicsNEXT LESSON →Finding Short Vectors in LatticesArticle · Engineering MathematicsLLL for Linearly Dependent Generating SetsArticle · Engineering MathematicsPolynomial Representation and StorageArticle · Engineering Mathematics
KEVOS · Engineering, manufacturing and project improvement
ArticlesServicesCase studiesAboutContact
© 2026 KEVOS®