Explicit sparse fractional cluster-LP solutions
DefinitionDualSepClusterLP_SparseInputapproximation-algorithmscluster-lpcorrelation-clustering
Encode a list of distinct subsets with positive rational weights. The induced primal vector is zero outside the list and must have exact unit vertex marginals.
Definition code
import Mathlib
import Mathlib.Computability.TuringMachine.Computable
import Definitions.Def_DualSepClusterLP_LinearProgram
import Definitions.Def_DualSepClusterLP_SeparationEncoding
namespace DualSepClusterLP
/-- A list of subsets and their explicitly represented rational LP weights. -/
abbrev SparseWeights (n : ℕ) := List (Finset (Fin n) × ℚ)
/-- The vector indexed by all subsets, extended by zero off the listed support. -/
def weightAt {n : ℕ} (L : SparseWeights n) (S : Finset (Fin n)) : ℚ :=
(L.map fun e => if e.1 = S then e.2 else 0).sum
/-- A sparse representation contains each support set once with its positive weight. -/
def ValidSupport {n : ℕ} (L : SparseWeights n) : Prop :=
(L.map Prod.fst).Nodup ∧ (∀ e ∈ L, 0 < e.2) ∧
PrimalFeasible (fun S => (weightAt L S : ℝ))
/-- Graph header (accuracy field zero), self-delimiting support length, then an n-bit
membership mask and a self-delimiting rational weight for each listed subset. -/
def roundingInput {n : ℕ} (G : Graph n) (L : SparseWeights n) : List Bool :=
graphInput 0 G ++ lengthPrefixedNat L.length ++
L.flatMap (fun e => (List.finRange n).map (fun i => decide (i ∈ e.1)) ++ priceBits e.2)
end DualSepClusterLP
Source
David García-Soriano and Antoine Schohn, Approximate Dual Separation for the Cluster LP, STOC 2027 paper #79, §7 and Appendix F, explicitly represented primal solution. https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/79.pdf
Read-back
What the Lean code literally says, in plain math · Codex (GPT-6)
The fractional solution is a list of distinct vertex subsets and strictly positive rational weights. An unlisted subset has weight zero. Validity requires the weights of the subsets containing each vertex to sum to one. A graph encoding, number of entries, and one n-bit mask and a rational field per entry form the rounding input.
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.