OctoInformBeta

OpenAI preprint, September 2026 · §4. Gaussian heat dissipation

§4 (setting): Gaussian weights, scores, quadratic costs, and the center plane: Densities and scores, heat-weighted star matrices on a finite set ι\iota, local Gram Hessian terms, plane integrals, and the cubic array norm

Gaussian weights and variational energies on center planes.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Discretizing a trilinear form on the real line produces a triple of complex matrices acting on a finite set of real nodes at three cyclic vertices. Localizing each vertex by a Gaussian kernel scales the matrix entries across edges and yields a vertex star matrix whose 3/23/2-power energy varies with the vertex centers and variances. Tracking those variations leads to score-weighted local Gram matrices and the Hessian quadratic cost of the matrix energy, while integrating over constant-sum center planes connects the total heat energy to a cubic array norm.

Matrix Energy and Gram Preliminaries

Let α\alpha be a finite index set, and write ℂα×α\mathbb{C}^{\alpha \times \alpha} for the space of complex square matrices indexed by α\alpha. For A∈ℂα×αA \in \mathbb{C}^{\alpha \times \alpha}, the real trace tr⁡ℝ(A)\operatorname{tr}_{\mathbb{R}}(A) is the real number Re⁡tr⁡(A)\operatorname{Re}\operatorname{tr}(A), and the real trace functional tr⁡ℝ\operatorname{tr}_{\mathbb{R}} on ℂα×α\mathbb{C}^{\alpha \times \alpha} is the continuous real-linear functional defined by A↦tr⁡ℝ(A)A \mapsto \operatorname{tr}_{\mathbb{R}}(A). Given X,Y∈ℂα×αX, Y \in \mathbb{C}^{\alpha \times \alpha}, the real Hilbert-Schmidt inner product ⟨X,Y⟩HS\langle X, Y \rangle_{\mathrm{HS}} is the real number tr⁡ℝ(X*Y)\operatorname{tr}_{\mathbb{R}}(X^* Y).

Given A∈ℂα×αA \in \mathbb{C}^{\alpha \times \alpha}, the Sylvester operator SAS_A on ℂα×α\mathbb{C}^{\alpha \times \alpha} is the linear operator mapping X∈ℂα×αX \in \mathbb{C}^{\alpha \times \alpha} to AX+XAA X + X A, and the Sylvester penalty of AA and XX is the real number ⟨X,SA(X)⟩HS\langle X, S_A(X) \rangle_{\mathrm{HS}}.✦ The variational functional ψA,K\psi_{A,K} on ℂα×α\mathbb{C}^{\alpha \times \alpha} associated with A,K∈ℂα×αA, K \in \mathbb{C}^{\alpha \times \alpha} is the real-valued function ψA,K(X)=2⟨K,X⟩HS−⟨X,SA(X)⟩HS\psi_{A,K}(X) = 2\langle K, X \rangle_{\mathrm{HS}} - \langle X, S_A(X) \rangle_{\mathrm{HS}}, and the root variational cost croot(A,K)c_{\mathrm{root}}(A, K) is the real number 32sup⁡X∈ℂα×αψA,K(X)\frac{3}{2} \sup_{X \in \mathbb{C}^{\alpha \times \alpha}} \psi_{A,K}(X).

With P1/2P^{1/2} denoting the square root of P∈ℂα×αP \in \mathbb{C}^{\alpha \times \alpha} defined by the continuous functional calculus, the matrix energy Φ(P)\Phi(P) (written tr⁡(P3/2)\operatorname{tr}(P^{3/2})) is the real number tr⁡ℝ(PP1/2)\operatorname{tr}_{\mathbb{R}}(P P^{1/2}). At P∈ℂα×αP \in \mathbb{C}^{\alpha \times \alpha}, the matrix energy gradient DΦ(P)D\Phi(P) is the continuous real-linear functional on ℂα×α\mathbb{C}^{\alpha \times \alpha} given for K∈ℂα×αK \in \mathbb{C}^{\alpha \times \alpha} by DΦ(P)(K)=32tr⁡ℝ(P1/2K)D\Phi(P)(K) = \frac{3}{2} \operatorname{tr}_{\mathbb{R}}(P^{1/2} K).

The Hessian quadratic cost bP(K)b_P(K) of P,K∈ℂα×αP, K \in \mathbb{C}^{\alpha \times \alpha} is the real number croot(P1/2,K)c_{\mathrm{root}}(P^{1/2}, K).✦ For P,K,L∈ℂα×αP, K, L \in \mathbb{C}^{\alpha \times \alpha}, the Hessian bilinear form ℬP(K,L)\mathcal{B}_P(K, L) is the real number 14(bP(K+L)−bP(K−L))\frac{1}{4}\bigl(b_P(K + L) - b_P(K - L)\bigr).

Given B∈ℂα×αB \in \mathbb{C}^{\alpha \times \alpha} and a real weight vector μ:α→ℝ\mu : \alpha \to \mathbb{R}, the weighted Gram matrix Γμ(B)\Gamma_\mu(B) is the matrix B*diag⁡(μ)BB^* \operatorname{diag}(\mu) B in ℂα×α\mathbb{C}^{\alpha \times \alpha}.

Triangle Block Matrices

Index the three vertices by v∈Fin⁡3v \in \operatorname{Fin} 3, equipped with cyclic addition v+1v + 1 and v−1=v+2v - 1 = v + 2. For a finite index set ι\iota, the triangle block space is Fin⁡3×ι\operatorname{Fin} 3 \times \iota, on which any 3×33 \times 3 array of matrices in ℂι×ι\mathbb{C}^{\iota \times \iota} assembles into a complex square matrix in ℂ(Fin⁡3×ι)×(Fin⁡3×ι)\mathbb{C}^{(\operatorname{Fin} 3 \times \iota) \times (\operatorname{Fin} 3 \times \iota)}.

Given a matrix triple W:Fin⁡3→ℂι×ιW : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, the triangle matrix Θ(W)\Theta(W) is the block matrix on Fin⁡3×ι\operatorname{Fin} 3 \times \iota whose (v,w)(v, w)-block is WvW_v when w=v+1w = v + 1, Wv−1*W_{v-1}^* when w=v−1w = v - 1, and 00 when w=vw = v. For each v∈Fin⁡3v \in \operatorname{Fin} 3, the vertex block projection QvQ_v is the diagonal matrix in ℂ(Fin⁡3×ι)×(Fin⁡3×ι)\mathbb{C}^{(\operatorname{Fin} 3 \times \iota) \times (\operatorname{Fin} 3 \times \iota)} equal to 11 on entries with first coordinate vv and 00 elsewhere. For W:Fin⁡3→ℂι×ιW : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} and v∈Fin⁡3v \in \operatorname{Fin} 3, the vertex star row Rv(W)R_v(W) is the block matrix QvΘ(W)Q_v \Theta(W) on Fin⁡3×ι\operatorname{Fin} 3 \times \iota, and the vertex star Gram matrix Tv(W)T_v(W) is the block matrix Rv(W)*Rv(W)R_v(W)^* R_v(W) on Fin⁡3×ι\operatorname{Fin} 3 \times \iota.

For a vertex weight family g:Fin⁡3→ι→ℝg : \operatorname{Fin} 3 \to \iota \to \mathbb{R}, the vertex root matrix D(g)D(g) is the diagonal matrix in ℂ(Fin⁡3×ι)×(Fin⁡3×ι)\mathbb{C}^{(\operatorname{Fin} 3 \times \iota) \times (\operatorname{Fin} 3 \times \iota)} with real square-root entry gz(i)\sqrt{g_z(i)} at (z,i)(z, i). Given A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} and g:Fin⁡3→ι→ℝg : \operatorname{Fin} 3 \to \iota \to \mathbb{R}, the weighted matrix array A(g):Fin⁡3→ℂι×ιA^{(g)} : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} is the triple with (v,i,j)(v, i, j) entry gv(i)Av(i,j)gv+1(j)\sqrt{g_v(i)}\, A_v(i, j)\, \sqrt{g_{v+1}(j)}. For A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, g:Fin⁡3→ι→ℝg : \operatorname{Fin} 3 \to \iota \to \mathbb{R}, and v∈Fin⁡3v \in \operatorname{Fin} 3, the vertex star base row Bvrow(A,g)B^{\mathrm{row}}_v(A, g) is the block matrix Rv(A)D(g)R_v(A) D(g) on Fin⁡3×ι\operatorname{Fin} 3 \times \iota, and the vertex star base column Bvcol(A,g)B^{\mathrm{col}}_v(A, g) is the block matrix Rv(A)*D(g)R_v(A)^* D(g) on Fin⁡3×ι\operatorname{Fin} 3 \times \iota.✦

The Definition

Definition 1. For real numbers σ,t∈ℝ\sigma, t \in \mathbb{R}, the Gaussian density gσ(t)g_\sigma(t) is the real number defined by (1): gσ(t)=(2πσ)−1/2exp⁡(−t22σ).g_\sigma(t) = (2\pi\sigma)^{-1/2} \exp\!\left(-\frac{t^2}{2\sigma}\right).(1)✦ Given h,u,p,σ∈ℝh, u, p, \sigma \in \mathbb{R}, the Gaussian weight w(h,u,p,σ)w(h, u, p, \sigma) is the real number (2): w(h,u,p,σ)=hgσ(u−p).w(h, u, p, \sigma) = h\, g_\sigma(u - p).(2)✦ The Gaussian score s(u,p,σ)s(u, p, \sigma) for u,p,σ∈ℝu, p, \sigma \in \mathbb{R} is the real number given by (3): s(u,p,σ)=u−pσ.s(u, p, \sigma) = \frac{u - p}{\sigma}.(3)✦ For sets α\alpha and ν\nu, an assignment ℓ:α→ν\ell : \alpha \to \nu, scalar h∈ℝh \in \mathbb{R}, coordinates x:α→ℝx : \alpha \to \mathbb{R}, and parameter functions p,σ:ν→ℝp, \sigma : \nu \to \mathbb{R}, the grouped heat weight μ(h,x,ℓ,p,σ):α→ℝ\mu(h, x, \ell, p, \sigma) : \alpha \to \mathbb{R} is the function mapping a∈αa \in \alpha to w(h,x(a),p(ℓ(a)),σ(ℓ(a)))w(h, x(a), p(\ell(a)), \sigma(\ell(a))). Given ℓ:α→ν\ell : \alpha \to \nu, h∈ℝh \in \mathbb{R}, x:α→ℝx : \alpha \to \mathbb{R}, p,σ:ν→ℝp, \sigma : \nu \to \mathbb{R}, and r∈νr \in \nu for sets α\alpha and ν\nu, the grouped first-variation weight μr(1)(h,x,ℓ,p,σ):α→ℝ\mu^{(1)}_r(h, x, \ell, p, \sigma) : \alpha \to \mathbb{R} is the function mapping a∈αa \in \alpha to s(x(a),p(r),σ(r))w(h,x(a),p(r),σ(r))s(x(a), p(r), \sigma(r))\, w(h, x(a), p(r), \sigma(r)) if ℓ(a)=r\ell(a) = r and to 00 otherwise, while the grouped second-variation weight μr(2)(h,x,ℓ,p,σ):α→ℝ\mu^{(2)}_r(h, x, \ell, p, \sigma) : \alpha \to \mathbb{R} is the function mapping a∈αa \in \alpha to (s(x(a),p(r),σ(r))2−σ(r)−1)w(h,x(a),p(r),σ(r))\bigl(s(x(a), p(r), \sigma(r))^2 - \sigma(r)^{-1}\bigr)\, w(h, x(a), p(r), \sigma(r)) if ℓ(a)=r\ell(a) = r and to 00 otherwise.

Fix a finite index set ι\iota. For h∈ℝh \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, and p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R}, the vertex heat weight family 𝒘(h,u,p,σ):Fin⁡3→ι→ℝ\mathbf{w}(h, u, p, \sigma) : \operatorname{Fin} 3 \to \iota \to \mathbb{R} assigns to v∈Fin⁡3v \in \operatorname{Fin} 3 and i∈ιi \in \iota the node weight wv(i)=𝒘(h,u,p,σ)v(i)w_v(i) = \mathbf{w}(h, u, p, \sigma)_v(i) defined in (4): wv(i)=w(h,ui,pv,σv).w_v(i) = w(h, u_i, p_v, \sigma_v).(4)✦ Given a matrix triple A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} together with h∈ℝh \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, and p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R}, the heat matrix array W(A,h,u,p,σ):Fin⁡3→ℂι×ιW(A, h, u, p, \sigma) : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} is the weighted matrix array (5): W(A,h,u,p,σ)=A(𝒘(h,u,p,σ)).W(A, h, u, p, \sigma) = A^{(\mathbf{w}(h, u, p, \sigma))}.(5)✦ At a vertex v∈Fin⁡3v \in \operatorname{Fin} 3, the heat matrix energy ev(A,h,u,p,σ)e_v(A, h, u, p, \sigma) for A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, h∈ℝh \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, and p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R} is the real number given by (6): ev(A,h,u,p,σ)=Φ(Tv(W(A,h,u,p,σ))).e_v(A, h, u, p, \sigma) = \Phi\bigl(T_v(W(A, h, u, p, \sigma))\bigr).(6)✦

For A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, vertex weights g:Fin⁡3→ι→ℝg : \operatorname{Fin} 3 \to \iota \to \mathbb{R}, and vertices v,w∈Fin⁡3v, w \in \operatorname{Fin} 3, the fixed base matrix Fv,w(A,g)F_{v,w}(A, g) is the block matrix on Fin⁡3×ι\operatorname{Fin} 3 \times \iota defined by (7): Fv,w(A,g)={Bvrow(A,g),w=v,Bvcol(A,g),w≠v.F_{v,w}(A, g) = \begin{cases} B^{\mathrm{row}}_v(A, g), & w = v, \\ B^{\mathrm{col}}_v(A, g), & w \neq v. \end{cases}(7)✦ The coordinate projection π1:Fin⁡3×ι→Fin⁡3\pi_1 : \operatorname{Fin} 3 \times \iota \to \operatorname{Fin} 3 is the map given by π1(z,i)=z\pi_1(z, i) = z, and for u:ι→ℝu : \iota \to \mathbb{R}, the extended coordinate function ũ:Fin⁡3×ι→ℝ\tilde u : \operatorname{Fin} 3 \times \iota \to \mathbb{R} is the function given by ũ(z,i)=ui\tilde u(z, i) = u_i. For A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, h∈ℝh \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R}, and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3, the local Gram matrix Pv,wP_{v,w} is the block matrix on Fin⁡3×ι\operatorname{Fin} 3 \times \iota given by (8): Pv,w=Γμ(h,ũ,π1,p,σ)(Fv,w(A,𝒘(h,u,p,σ))).P_{v,w} = \Gamma_{\mu(h, \tilde u, \pi_1, p, \sigma)}\bigl(F_{v,w}(A, \mathbf{w}(h, u, p, \sigma))\bigr).(8)✦ The local first variation Kv,w(1)K^{(1)}_{v,w} for (A,h,u,p,σ)(A, h, u, p, \sigma) and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3 is the block matrix on Fin⁡3×ι\operatorname{Fin} 3 \times \iota in (9): Kv,w(1)=Γμw(1)(h,ũ,π1,p,σ)(Fv,w(A,𝒘(h,u,p,σ))).K^{(1)}_{v,w} = \Gamma_{\mu^{(1)}_w(h, \tilde u, \pi_1, p, \sigma)}\bigl(F_{v,w}(A, \mathbf{w}(h, u, p, \sigma))\bigr).(9)✦ At (A,h,u,p,σ)(A, h, u, p, \sigma) and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3, the local second variation Kv,w(2)K^{(2)}_{v,w} is the block matrix on Fin⁡3×ι\operatorname{Fin} 3 \times \iota defined by (10): Kv,w(2)=Γμw(2)(h,ũ,π1,p,σ)(Fv,w(A,𝒘(h,u,p,σ))).K^{(2)}_{v,w} = \Gamma_{\mu^{(2)}_w(h, \tilde u, \pi_1, p, \sigma)}\bigl(F_{v,w}(A, \mathbf{w}(h, u, p, \sigma))\bigr).(10)✦

For A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, h∈ℝh \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R}, and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3, with Pv,wP_{v,w}, Kv,w(1)K^{(1)}_{v,w}, and Kv,w(2)K^{(2)}_{v,w} as in (8), (9), and (10), the heat energy gradient ℰv,w(1)\mathcal{E}^{(1)}_{v,w} is the real number (11): ℰv,w(1)=DΦ(Pv,w)(Kv,w(1)).\mathcal{E}^{(1)}_{v,w} = D\Phi(P_{v,w})\bigl(K^{(1)}_{v,w}\bigr).(11)✦ The pure heat second variation ℰv,wpure\mathcal{E}^{\mathrm{pure}}_{v,w} at (A,h,u,p,σ)(A, h, u, p, \sigma) and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3 is the real number given by (12): ℰv,wpure=bPv,w(Kv,w(1))+DΦ(Pv,w)(Kv,w(2)).\mathcal{E}^{\mathrm{pure}}_{v,w} = b_{P_{v,w}}\bigl(K^{(1)}_{v,w}\bigr) + D\Phi(P_{v,w})\bigl(K^{(2)}_{v,w}\bigr).(12)✦ At (A,h,u,p,σ)(A, h, u, p, \sigma) and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3, the heat variance term ℰv,wvar\mathcal{E}^{\mathrm{var}}_{v,w} is the real number (13): ℰv,wvar=12DΦ(Pv,w)(Kv,w(2)).\mathcal{E}^{\mathrm{var}}_{v,w} = \frac{1}{2}\, D\Phi(P_{v,w})\bigl(K^{(2)}_{v,w}\bigr).(13)✦ The heat Hessian cost ℰv,wcost\mathcal{E}^{\mathrm{cost}}_{v,w} for (A,h,u,p,σ)(A, h, u, p, \sigma) and v,w∈Fin⁡3v, w \in \operatorname{Fin} 3 is the real number defined by (14): ℰv,wcost=bPv,w(Kv,w(1)).\mathcal{E}^{\mathrm{cost}}_{v,w} = b_{P_{v,w}}\bigl(K^{(1)}_{v,w}\bigr).(14)✦ Given (A,h,u,p,σ)(A, h, u, p, \sigma) and three vertices v,w,z∈Fin⁡3v, w, z \in \operatorname{Fin} 3, the mixed heat Hessian pairing ℰv,w,zmix\mathcal{E}^{\mathrm{mix}}_{v,w,z} is the real number (15): ℰv,w,zmix=ℬPv,w(Kv,w(1),Kv,z(1)).\mathcal{E}^{\mathrm{mix}}_{v,w,z} = \mathcal{B}_{P_{v,w}}\bigl(K^{(1)}_{v,w}, K^{(1)}_{v,z}\bigr).(15)✦

For real level d∈ℝd \in \mathbb{R}, the constant-sum center plane Πd\Pi_d is the plane {p0+p1+p2=d}\{p_0 + p_1 + p_2 = d\}, and for an omitted coordinate i∈Fin⁡3i \in \operatorname{Fin} 3 and z=(z1,z2)∈ℝ2z = (z_1, z_2) \in \mathbb{R}^2, the plane coordinate chart γi(d,z)\gamma_i(d, z) is the point in ℝFin⁡3\mathbb{R}^{\operatorname{Fin} 3} defined by (16): γi(d,z)={(d−z1−z2,z1,z2),i=0,(z1,d−z1−z2,z2),i=1,(z1,z2,d−z1−z2),i=2.\gamma_i(d, z) = \begin{cases} (d - z_1 - z_2,\, z_1,\, z_2), & i = 0, \\ (z_1,\, d - z_1 - z_2,\, z_2), & i = 1, \\ (z_1,\, z_2,\, d - z_1 - z_2), & i = 2. \end{cases}(16)✦ Given a function f:ℝFin⁡3→ℝf : \mathbb{R}^{\operatorname{Fin} 3} \to \mathbb{R} and d∈ℝd \in \mathbb{R}, the plane integral ℐd(f)\mathcal{I}_d(f) is the real number defined using the chart i=2i = 2 by (17): ℐd(f)=∫ℝ2f(γ2(d,z))dz.\mathcal{I}_d(f) = \int_{\mathbb{R}^2} f\bigl(\gamma_2(d, z)\bigr)\, dz.(17)✦ The total plane heat energy ℰtot(A,h,u,d,σ)\mathcal{E}_{\mathrm{tot}}(A, h, u, d, \sigma) for A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, h,d∈ℝh, d \in \mathbb{R}, u:ι→ℝu : \iota \to \mathbb{R}, and σ:Fin⁡3→ℝ\sigma : \operatorname{Fin} 3 \to \mathbb{R} is the real number (18): ℰtot(A,h,u,d,σ)=∑v∈Fin⁡3ℐd(p↦ev(A,h,u,p,σ)).\mathcal{E}_{\mathrm{tot}}(A, h, u, d, \sigma) = \sum_{v \in \operatorname{Fin} 3} \mathcal{I}_d\bigl(p \mapsto e_v(A, h, u, p, \sigma)\bigr).(18)✦ For a matrix triple A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota} and step size h∈ℝh \in \mathbb{R}, the cubic array norm 𝒮(A)\mathcal{S}(A) is the real number (19): 𝒮(A)=∑v∈Fin⁡3∑i,j∈ιh2|Av(i,j)|3.\mathcal{S}(A) = \sum_{v \in \operatorname{Fin} 3} \sum_{i, j \in \iota} h^2 |A_v(i, j)|^3.(19)✦

In the paper, the result reads: for h>0h > 0, integer N≥0N \ge 0, nodes ui=ihu_i = ih (−N≤i≤N-N \le i \le N), arrays Av(i,j)A_v(i, j), centers pv∈ℝp_v \in \mathbb{R}, and variances σv>0\sigma_v > 0, one defines gσ(t)=(2πσ)−1/2exp⁡(−t2/(2σ))g_\sigma(t) = (2\pi\sigma)^{-1/2} \exp(-t^2/(2\sigma)), wv(i)=hgσv(ui−pv)w_v(i) = h g_{\sigma_v}(u_i - p_v), Wv(i,j)=Av(i,j)wv(i)wv+1(j)W_v(i, j) = A_v(i, j) \sqrt{w_v(i) w_{v+1}(j)}, ev=Φ(Tv)e_v = \Phi(T_v), Lw=diag⁡i((ui−pw)/σw)L_w = \operatorname{diag}_i((u_i - p_w)/\sigma_w), Uv=Rv*LvRvU_v = R_v^* L_v R_v, Uv,w=W*LwW**U_{v,w} = W_* L_w W_*^* (w≠vw \neq v), Vv=bTv(Uv)V_v = b_{T_v}(U_v), Yv,w=bSv(Uv,w)Y_{v,w} = b_{S_v}(U_{v,w}), Zv=ℬSv(Uv,v+1,Uv,v−1)Z_v = \mathcal{B}_{S_v}(U_{v,v+1}, U_{v,v-1}), the plane Πd={p0+p1+p2=d}\Pi_d = \{p_0 + p_1 + p_2 = d\}, λ=11/10\lambda = 11/10, and 𝒮(A)=∑v∑i,jh2|Av(i,j)|3\mathcal{S}(A) = \sum_v \sum_{i,j} h^2 |A_v(i, j)|^3, where Tv(W)T_v(W), Rv(W)R_v(W), ev(A,h,u,p,σ)e_v(A, h, u, p, \sigma), Pv,wP_{v,w} (w≠vw \neq v), Kv,v(1)K^{(1)}_{v,v}, Kv,w(1)K^{(1)}_{v,w} (w≠vw \neq v), ℰv,vcost\mathcal{E}^{\mathrm{cost}}_{v,v}, ℰv,wcost\mathcal{E}^{\mathrm{cost}}_{v,w} (w≠vw \neq v), and ℰv,v+1,v−1mix\mathcal{E}^{\mathrm{mix}}_{v,v+1,v-1} stand for TvT_v, RvR_v, eve_v, SvS_v, UvU_v, Uv,wU_{v,w}, VvV_v, Yv,wY_{v,w}, and ZvZ_v. By contrast, a general finite index set ι\iota, arbitrary node coordinates u:ι→ℝu : \iota \to \mathbb{R}, vertex indices v∈Fin⁡3v \in \operatorname{Fin} 3, and unconstrained real parameters hh, pp, and σ\sigma replace the symmetric lattice {−N,…,N}\{-N, \dots, N\} and positive variances, while the fixed base matrix and local Gram matrix unify the diagonal (w=vw = v) and off-diagonal (w≠vw \neq v) cases.

What Is Checked

The definitions are stated over an arbitrary finite index set ι\iota, complex square matrix triples A:Fin⁡3→ℂι×ιA : \operatorname{Fin} 3 \to \mathbb{C}^{\iota \times \iota}, real node coordinates u:ι→ℝu : \iota \to \mathbb{R}, arbitrary real scalars h,d,t∈ℝh, d, t \in \mathbb{R}, real vertex triples p,σ:Fin⁡3→ℝp, \sigma : \operatorname{Fin} 3 \to \mathbb{R}, vertex weights g:Fin⁡3→ι→ℝg : \operatorname{Fin} 3 \to \iota \to \mathbb{R}, and real-valued functions f:ℝFin⁡3→ℝf : \mathbb{R}^{\operatorname{Fin} 3} \to \mathbb{R}, together with arbitrary sets α\alpha and ν\nu for the grouped weights. For the underlying constructions, the library proves that the real trace on complex square matrices indexed by a finite set is additive and real-homogeneous.