Documentation

LeanPool.BollobasNikiforov.Spectral.Conic

The rank-two conic parameter χ''_{vec,3} #

chiVec3 G is the supremum of the Frobenius mass inner X X over PSD matrices of rank at most two, nonnegative on the edges of G, and normalized so that the mass of X ⊙ X off those edges equals 1. The feasible set is nonempty and compact, so the supremum is attained.

The all-ones matrix J.

Equations
Instances For

    Off-edge indicator I + A_{Gᶜ} = J - A_G. Equals 1 on the diagonal and on non-edges of G, and 0 on edges.

    Equations
    Instances For
      theorem BollobasNikiforov.onesMatrix_apply {V : Type u_1} (i j : V) :
      theorem BollobasNikiforov.inner_offEdgeMatrix_hadamard_self {V : Type u_1} [Fintype V] [DecidableEq V] {G : SimpleGraph V} [DecidableRel G.Adj] (X : Matrix V V ℝ) :
      inner (offEdgeMatrix G) (X.hadamard X) = ∑ i : V, ∑ j : V, if G.Adj i j then 0 else X i j ^ 2

      The mass on the diagonal is at most the off-edge normalization.

      CG06 — feasible set and the parameter #

      Feasible matrices for χ''_{vec,3}.

      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        noncomputable def BollobasNikiforov.chiVec3 {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] :

        CG06. The parameter χ''_{vec,3}(G).

        Equations
        Instances For

          Rank-two writing #

          theorem BollobasNikiforov.exists_add_vecMulVec_of_posSemidef_rank_le_two {V : Type u_1} [Fintype V] {X : Matrix V V ℝ} (hX : X.PosSemidef) (hr : X.rank ≤ 2) :
          ∃ (u : V → ℝ) (v : V → ℝ), X = Matrix.vecMulVec u u + Matrix.vecMulVec v v

          A real PSD matrix of rank at most two is a sum of two real outer products.

          CG07 — nonempty compact feasible set #

          def BollobasNikiforov.chiVec3Param {V : Type u_1} [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj] :
          Set ((V → ℝ) × (V → ℝ))

          Parameter domain for the rank-two writing: bounded outer-product data satisfying the normalization and edge-nonnegativity constraints.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            theorem BollobasNikiforov.continuous_add_vecMulVec {V : Type u_1} :
            Continuous fun (p : (V → ℝ) × (V → ℝ)) => Matrix.vecMulVec p.1 p.1 + Matrix.vecMulVec p.2 p.2
            theorem BollobasNikiforov.mem_pi_Icc_iff {V : Type u_1} {u : V → ℝ} :
            (u ∈ Set.univ.pi fun (x : V) => Set.Icc (-1) 1) ↔ ∀ (i : V), u i ∈ Set.Icc (-1) 1
            theorem BollobasNikiforov.mem_chiVec3Param_of_add_vecMulVec {V : Type u_1} [Fintype V] [DecidableEq V] {G : SimpleGraph V} [DecidableRel G.Adj] {X : Matrix V V ℝ} {u v : V → ℝ} (hX : X = Matrix.vecMulVec u u + Matrix.vecMulVec v v) (hnorm : inner (offEdgeMatrix G) (X.hadamard X) = 1) (hnn : ∀ (i j : V), G.Adj i j → 0 ≤ X i j) :
            theorem BollobasNikiforov.ChiVec3Feasible.mem_image_param {V : Type u_1} [Fintype V] [DecidableEq V] {G : SimpleGraph V} [DecidableRel G.Adj] {X : Matrix V V ℝ} (hX : ChiVec3Feasible G X) :
            X ∈ (fun (p : (V → ℝ) × (V → ℝ)) => Matrix.vecMulVec p.1 p.1 + Matrix.vecMulVec p.2 p.2) '' chiVec3Param G
            theorem BollobasNikiforov.ChiVec3Feasible.of_mem_image_param {V : Type u_1} [Fintype V] [DecidableEq V] {G : SimpleGraph V} [DecidableRel G.Adj] {X : Matrix V V ℝ} (hX : X ∈ (fun (p : (V → ℝ) × (V → ℝ)) => Matrix.vecMulVec p.1 p.1 + Matrix.vecMulVec p.2 p.2) '' chiVec3Param G) :

            CG07. A standard-basis rank-one matrix is feasible after the normalization (already equal to 1).

            CG07. The feasible set is compact.

            CG07. The Frobenius objective attains its maximum on the feasible set.

            CG09 — clique Gram matrix #

            Indicator of a vertex set.

            Equations
            Instances For
              noncomputable def BollobasNikiforov.cliqueGram {V : Type u_1} [DecidableEq V] (s : Finset V) :

              Normalized rank-one Gram matrix of a clique: qqᵀ / √r.

              Equations
              Instances For
                theorem BollobasNikiforov.cliqueGram_apply {V : Type u_1} [DecidableEq V] (s : Finset V) (i j : V) :
                cliqueGram s i j = if i ∈ s ∧ j ∈ s then 1 / √↑s.card else 0
                theorem BollobasNikiforov.cliqueGram_sq_apply {V : Type u_1} [DecidableEq V] {s : Finset V} (hs : s.Nonempty) (i j : V) :
                cliqueGram s i j ^ 2 = if i ∈ s ∧ j ∈ s then 1 / ↑s.card else 0
                theorem BollobasNikiforov.rank_smul_le {V : Type u_1} [Fintype V] (c : ℝ) (A : Matrix V V ℝ) :
                (c • A).rank ≤ A.rank

                CG09. Off-edge mass of the clique Gram matrix is 1.

                CG09. The clique Gram matrix is feasible.

                theorem BollobasNikiforov.inner_cliqueGram {V : Type u_1} [Fintype V] [DecidableEq V] {s : Finset V} (hne : s.Nonempty) :

                CG09. The clique Gram matrix has Frobenius mass s.card.

                A maximum clique witnesses ω(G) ≤ χ''_{vec,3}(G).

                CG08 — upper bound chiVec3 ≤ cliqueNum #

                theorem BollobasNikiforov.inner_adjMatrix_hadamard_self_eq_sum_posPart_sq {V : Type u_1} [Fintype V] {G : SimpleGraph V} [DecidableRel G.Adj] {X : Matrix V V ℝ} (hnn : ∀ (i j : V), G.Adj i j → 0 ≤ X i j) :
                inner (SimpleGraph.adjMatrix ℝ G) (X.hadamard X) = ∑ i : V, ∑ j : V, SimpleGraph.adjMatrix ℝ G i j * posPart X i j ^ 2

                A feasible matrix has Frobenius mass at most ω(G).

                CG08. Every feasible objective is ≤ ω(G), so the supremum is too.

                CG10 — cor:parameter #

                CG10. Corollary cor:parameter.