TheoremBase

Proof of Localisation at a Sequentially Strict Maximum of a Quadratically Penalised Difference

lemmalem:penalised-difference-localisation-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 9,685 chars · 12 deps · depth 22 Reason: Proof of the localisation lemma: near-maximising pairs converge to the maximum point by sequential strictness, the one-sided value bounds follow from the closed superlevel sets of the two functions, and the remaining two by solving for one value in terms of the other.

Four reductions: near-maximising pairs converge to the maximum point by sequential strictness; the one-sided value bounds come from the closed superlevel sets of the two functions; the remaining two follow by solving for one value in terms of the other.

Proof

Each result cited is universally quantified over the data in its own statement, and is applied here to the data named in the statement of the lemma.

Write M0=Φ(xˉ,yˉ)M_{0}=\Phi(\bar{x},\bar{y}) and R=xˉyˉR=|\bar{x}-\bar{y}|, so that 0R0\le R by Real Inner Product Space §norm. By clause 1 of Sequentially Strict Maxima and Minima on a Subset of a Metric Space §maximum,

Φ(x,y)M0for every (x,y)A×A.\Phi(x,y)\le M_{0}\qquad\text{for every }(x,y)\in A\times A .

Fix, once and for all, a sequence (εk)kN(\varepsilon_{k})_{k\in\mathbb{N}} of positive real numbers converging to 00 in (R,dR)(\mathbb{R},d_{\mathbb{R}}); such a sequence exists by Existence of a Sequence of Positive Real Numbers with Limit Zero. For x,yEx,y\in E one has d(x,y)=xyd(x,y)=|x-y| by Real Inner Product Space §distance, so that convergence of a sequence (xk)(x_{k}) to xx in (E,d)(E,d) means: for every positive σR\sigma\in\mathbb{R} there is k0Nk_{0}\in\mathbb{N} with xkx<σ|x_{k}-x|<\sigma for every kk with k0kk_{0}\le k.

Claim 1 (near-maximising pairs converge). Let (xk,yk)kN(x_{k},y_{k})_{k\in\mathbb{N}} be a sequence in A×AA\times A with M0εk<Φ(xk,yk)M_{0}-\varepsilon_{k}<\Phi(x_{k},y_{k}) for every kNk\in\mathbb{N}. Then (xk)kN(x_{k})_{k\in\mathbb{N}} converges to xˉ\bar{x} and (yk)kN(y_{k})_{k\in\mathbb{N}} converges to yˉ\bar{y} in (E,d)(E,d).

Proof. For each kk we have M0εk<Φ(xk,yk)M0M_{0}-\varepsilon_{k}<\Phi(x_{k},y_{k})\le M_{0}, hence 0M0Φ(xk,yk)<εk0\le M_{0}-\Phi(x_{k},y_{k})<\varepsilon_{k} by claim 3 of Elementary Arithmetic in an Ordered Field, and therefore Φ(xk,yk)M0εk|\Phi(x_{k},y_{k})-M_{0}|\le\varepsilon_{k} by claims 1 and 2 of Properties of the Absolute Value in an Ordered Field. Since (εk)(\varepsilon_{k}) converges to 00, claim 3 of Order Properties of Limits of Real Sequences shows that the sequence of real numbers (Φ(xk,yk))kN(\Phi(x_{k},y_{k}))_{k\in\mathbb{N}} converges to M0=Φ(xˉ,yˉ)M_{0}=\Phi(\bar{x},\bar{y}). The function Φ\Phi attains a sequentially strict maximum on A×AA\times A at (xˉ,yˉ)(\bar{x},\bar{y}), so clause 1 of Sequentially Strict Maxima and Minima on a Subset of a Metric Space §maximum gives that the sequence ((xk,yk))kN((x_{k},y_{k}))_{k\in\mathbb{N}} converges to (xˉ,yˉ)(\bar{x},\bar{y}) in (E×E,d×)(E\times E,d_{\times}). The coordinate maps of Real Hilbert Spaces: Series, Products, Orthonormal Bases and Differential Calculus §product send (xk,yk)(x_{k},y_{k}) to xkx_{k} and to yky_{k} and (xˉ,yˉ)(\bar{x},\bar{y}) to xˉ\bar{x} and to yˉ\bar{y}, so claim 4 of Properties of the Product of Two Real Inner Product Spaces gives the assertion. This proves Claim 1.

Claim 2 (localisation of the points). For every positive εR\varepsilon\in\mathbb{R} there is a positive ηR\eta\in\mathbb{R} such that every (x,y)A×A(x,y)\in A\times A with M0η<Φ(x,y)M_{0}-\eta<\Phi(x,y) satisfies xxˉ<ε|x-\bar{x}|<\varepsilon and yyˉ<ε|y-\bar{y}|<\varepsilon.

Proof. Suppose the assertion fails for some positive ε\varepsilon. Then for each kNk\in\mathbb{N}, failure for the positive number εk\varepsilon_{k} provides (xk,yk)A×A(x_{k},y_{k})\in A\times A with M0εk<Φ(xk,yk)M_{0}-\varepsilon_{k}<\Phi(x_{k},y_{k}) for which at least one of xkxˉ<ε|x_{k}-\bar{x}|<\varepsilon and ykyˉ<ε|y_{k}-\bar{y}|<\varepsilon fails, that is, for which εxkxˉ\varepsilon\le|x_{k}-\bar{x}| or εykyˉ\varepsilon\le|y_{k}-\bar{y}|. By Claim 1 the sequences (xk)(x_{k}) and (yk)(y_{k}) converge to xˉ\bar{x} and to yˉ\bar{y} in (E,d)(E,d), so there is k0Nk_{0}\in\mathbb{N} such that xkxˉ<ε|x_{k}-\bar{x}|<\varepsilon and ykyˉ<ε|y_{k}-\bar{y}|<\varepsilon both hold for every kk with k0kk_{0}\le k (take the larger of the two thresholds obtained for σ=ε\sigma=\varepsilon, using claim 9 of Elementary Order Arithmetic in an Ordered Field on the corresponding indices). For k=k0k=k_{0} this contradicts the previous sentence. This proves Claim 2.

Claim 3 (one-sided value bounds). For every positive εR\varepsilon\in\mathbb{R} there is a positive ηR\eta\in\mathbb{R} such that every (x,y)A×A(x,y)\in A\times A with M0η<Φ(x,y)M_{0}-\eta<\Phi(x,y) satisfies

u(x)<u(xˉ)+εandv(yˉ)ε<v(y).u(x)<u(\bar{x})+\varepsilon\qquad\text{and}\qquad v(\bar{y})-\varepsilon<v(y).

Proof. Suppose the first inequality fails to hold for all such pairs, for every positive η\eta. Then for each kNk\in\mathbb{N} there is (xk,yk)A×A(x_{k},y_{k})\in A\times A with M0εk<Φ(xk,yk)M_{0}-\varepsilon_{k}<\Phi(x_{k},y_{k}) and u(xˉ)+εu(xk)u(\bar{x})+\varepsilon\le u(x_{k}). By Claim 1 the sequence (xk)(x_{k}) lies in AA and converges to xˉ\bar{x} in (E,d)(E,d). The real number t=u(xˉ)+εt=u(\bar{x})+\varepsilon satisfies tu(xk)t\le u(x_{k}) for every kk, and uu has closed superlevel sets in EE, so claim 1 of Functions with Closed Superlevel Sets: Sequential Characterisation, Semicontinuity, Perturbation and Limits gives tu(xˉ)t\le u(\bar{x}), that is ε0\varepsilon\le0 by claim 3 of Elementary Arithmetic in an Ordered Field, contradicting 0<ε0<\varepsilon. Hence some positive η\eta' makes the first inequality hold for all pairs as described.

The second inequality is the first one applied to v-v and yˉ\bar{y}: if for every positive η\eta some pair (x,y)A×A(x,y)\in A\times A with M0η<Φ(x,y)M_{0}-\eta<\Phi(x,y) had v(y)v(yˉ)εv(y)\le v(\bar{y})-\varepsilon, that is v(yˉ)+εv(y)-v(\bar{y})+\varepsilon\le-v(y), then choosing such a pair (xk,yk)(x_{k},y_{k}) for η=εk\eta=\varepsilon_{k} and using that (yk)(y_{k}) converges to yˉ\bar{y} by Claim 1, that v-v has closed superlevel sets in EE, and claim 1 of Functions with Closed Superlevel Sets: Sequential Characterisation, Semicontinuity, Perturbation and Limits would give v(yˉ)+εv(yˉ)-v(\bar{y})+\varepsilon\le-v(\bar{y}), again contradicting 0<ε0<\varepsilon. Hence some positive η\eta'' makes the second inequality hold. Taking for η\eta the lesser of η\eta' and η\eta'', positive because it is one of them (claim 9 of Elementary Order Arithmetic in an Ordered Field), proves Claim 3.

Claim 4 (stability of the penalty). Let σR\sigma\in\mathbb{R} satisfy 0<σ10<\sigma\le1 and let x,yEx,y\in E satisfy xxˉ<σ|x-\bar{x}|<\sigma and yyˉ<σ|y-\bar{y}|<\sigma. Then

α2xy2α2R2  2ασ(R+1),\Bigl|\tfrac{\alpha}{2}|x-y|^{2}-\tfrac{\alpha}{2}R^{2}\Bigr|\ \le\ 2\,|\alpha|\,\sigma\,(R+1),

where R2=RRR^{2}=R\,R.

Proof. By the triangle inequality for the norm (claim 1 of The Norm Metric of a Real Inner Product Space: Triangle Inequalities, Limits and Continuity), applied twice in each line,

xyxxˉ+xˉyˉ+yˉy<R+2σ,Rxˉx+xy+yyˉ<xy+2σ,|x-y|\le|x-\bar{x}|+|\bar{x}-\bar{y}|+|\bar{y}-y|<R+2\sigma, \qquad R\le|\bar{x}-x|+|x-y|+|y-\bar{y}|<|x-y|+2\sigma,

so 2σ<xyR<2σ-2\sigma<|x-y|-R<2\sigma and hence xyR2σ\bigl||x-y|-R\bigr|\le2\sigma by claim 6 of Properties of the Absolute Value in an Ordered Field. Since σ1\sigma\le1 we also have 0xy+R<R+2σ+R2R+20\le|x-y|+R<R+2\sigma+R\le2R+2. Now xy2R2=(xyR)(xy+R)|x-y|^{2}-R^{2}=\bigl(|x-y|-R\bigr)\bigl(|x-y|+R\bigr), so by claim 4 of Properties of the Absolute Value in an Ordered Field and claim 5 of Elementary Arithmetic in an Ordered Field,

xy2R2=xyR(xy+R)2σ(2R+2)=4σ(R+1).\bigl||x-y|^{2}-R^{2}\bigr|=\bigl||x-y|-R\bigr|\,\bigl(|x-y|+R\bigr)\le 2\sigma\,(2R+2)=4\sigma\,(R+1).

Multiplying by the nonnegative number α2\tfrac{|\alpha|}{2} and using claim 4 of Properties of the Absolute Value in an Ordered Field again gives the claim. This proves Claim 4.

Claim 5 (the remaining value bounds). For every positive εR\varepsilon\in\mathbb{R} there is a positive ηR\eta\in\mathbb{R} such that every (x,y)A×A(x,y)\in A\times A with M0η<Φ(x,y)M_{0}-\eta<\Phi(x,y) satisfies

u(xˉ)ε<u(x)andv(y)<v(yˉ)+ε.u(\bar{x})-\varepsilon<u(x)\qquad\text{and}\qquad v(y)<v(\bar{y})+\varepsilon .

Proof. Let ε\varepsilon be positive. Choose a positive σ1\sigma\le1 with 2ασ(R+1)<ε42|\alpha|\sigma(R+1)<\tfrac{\varepsilon}{4}, as follows: if α=0\alpha=0 take σ=1\sigma=1, so that the left-hand side is 00; otherwise α|\alpha| is positive by claim 1 of Properties of the Absolute Value in an Ordered Field and R+1R+1 is positive, so the quotient ε16α(R+1)\tfrac{\varepsilon}{16|\alpha|(R+1)} exists and is positive by claim 7 of Elementary Order Arithmetic in an Ordered Field, and taking for σ\sigma the lesser of it and 11 gives 2ασ(R+1)ε8<ε42|\alpha|\sigma(R+1)\le\tfrac{\varepsilon}{8}<\tfrac{\varepsilon}{4}.

Let η1\eta_{1} be a number provided by Claim 2 for σ\sigma, let η2\eta_{2} be one provided by Claim 3 for ε2\tfrac{\varepsilon}{2}, and let η\eta be the least of η1\eta_{1}, η2\eta_{2} and ε4\tfrac{\varepsilon}{4}, positive because it is one of them.

Let (x,y)A×A(x,y)\in A\times A satisfy M0η<Φ(x,y)M_{0}-\eta<\Phi(x,y). Then xxˉ<σ|x-\bar{x}|<\sigma and yyˉ<σ|y-\bar{y}|<\sigma by Claim 2, so Claim 4 and claim 6 of Properties of the Absolute Value in an Ordered Field give

α2R2ε4<α2xy2;\tfrac{\alpha}{2}R^{2}-\tfrac{\varepsilon}{4}<\tfrac{\alpha}{2}|x-y|^{2};

and Claim 3 gives u(x)<u(xˉ)+ε2u(x)<u(\bar{x})+\tfrac{\varepsilon}{2} and v(yˉ)ε2<v(y)v(\bar{y})-\tfrac{\varepsilon}{2}<v(y). From the definition of Φ\Phi,

u(x)=Φ(x,y)+v(y)+α2xy2>(M0ε4)+(v(yˉ)ε2)+(α2R2ε4)=u(xˉ)ε,u(x)=\Phi(x,y)+v(y)+\tfrac{\alpha}{2}|x-y|^{2} >\Bigl(M_{0}-\tfrac{\varepsilon}{4}\Bigr)+\Bigl(v(\bar{y})-\tfrac{\varepsilon}{2}\Bigr)+\Bigl(\tfrac{\alpha}{2}R^{2}-\tfrac{\varepsilon}{4}\Bigr) =u(\bar{x})-\varepsilon,

the last equality because M0=u(xˉ)v(yˉ)α2R2M_{0}=u(\bar{x})-v(\bar{y})-\tfrac{\alpha}{2}R^{2}. Likewise

v(y)=Φ(x,y)u(x)+α2xy2>(M0ε4)(u(xˉ)+ε2)+(α2R2ε4)=v(yˉ)ε,-v(y)=\Phi(x,y)-u(x)+\tfrac{\alpha}{2}|x-y|^{2} >\Bigl(M_{0}-\tfrac{\varepsilon}{4}\Bigr)-\Bigl(u(\bar{x})+\tfrac{\varepsilon}{2}\Bigr)+\Bigl(\tfrac{\alpha}{2}R^{2}-\tfrac{\varepsilon}{4}\Bigr) =-v(\bar{y})-\varepsilon,

that is v(y)<v(yˉ)+εv(y)<v(\bar{y})+\varepsilon by claim 4 of Elementary Order Arithmetic in an Ordered Field. This proves Claim 5.

Conclusion. Let εR\varepsilon\in\mathbb{R} be positive. Let η1\eta_{1}, η2\eta_{2} and η3\eta_{3} be numbers provided for ε\varepsilon by Claims 2, 3 and 5 respectively, and let η\eta be the least of the three, positive because it is one of them. Let (x,y)A×A(x,y)\in A\times A satisfy Φ(xˉ,yˉ)η<Φ(x,y)\Phi(\bar{x},\bar{y})-\eta<\Phi(x,y). Claim 2 gives xxˉ<ε|x-\bar{x}|<\varepsilon and yyˉ<ε|y-\bar{y}|<\varepsilon. Claims 3 and 5 give

u(xˉ)ε<u(x)<u(xˉ)+ε,v(yˉ)ε<v(y)<v(yˉ)+ε,u(\bar{x})-\varepsilon<u(x)<u(\bar{x})+\varepsilon, \qquad v(\bar{y})-\varepsilon<v(y)<v(\bar{y})+\varepsilon,

that is ε<u(x)u(xˉ)<ε-\varepsilon<u(x)-u(\bar{x})<\varepsilon and ε<v(y)v(yˉ)<ε-\varepsilon<v(y)-v(\bar{y})<\varepsilon by claim 3 of Elementary Arithmetic in an Ordered Field, whence u(x)u(xˉ)<ε|u(x)-u(\bar{x})|<\varepsilon and v(y)v(yˉ)<ε|v(y)-v(\bar{y})|<\varepsilon by claim 9 of Properties of the Absolute Value in an Ordered Field. This is the assertion of the lemma.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…