Verified by 0 users · Statement flagged by 0 users
A second-order equation operator on the lift takes a square-integrable random vector whose law lies in a given set of measures, a real number, a second random vector, and a symmetric matrix, and depends on the two random vectors only through their joint law. Its delta-shifts relative to a penalty pair add or subtract delta times the penalty in the real slot and delta times the lifted score in the second random vector.
1. (Second-order equation operator on the lift) For a subset Q⊆P2(Rd), a second-order equation operator on the lift over Q is a function
F:QΛ×R×L2(Ω;Rd)×S(d)→R,
whose value at (X,r,V,X) is written F(X,r,V,X), with the following property. (Law-invariance in the pair) For all X,X′∈QΛ, all V,V′∈L2(Ω;Rd), every r∈R and every X∈S(d),
L(X,V)=L(X′,V′)impliesF(X,r,V,X)=F(X′,r,V′,X).
2. (The δ-shifts of F) Let F be a second-order equation operator on the lift over DΣ and let δ∈R be positive. For X∈DΣΛ write ν=L(X), an element of DΣ and hence of D by Penalty Pairs on the Wasserstein Space: the Penalty, Its Score, and Their Domains (i), so that E(ν) is a real number and the score Σ(ν) lies in Tν, hence in the space L2(ν;Rd), so that the compositionΣ(ν)∘X is an element of L2(Ω;Rd). The δ-shifts of F relative to the penalty pair are the two functions
Both are second-order equation operators on the lift over DΣ. Indeed, let X′∈DΣΛ and V′∈L2(Ω;Rd) satisfy L(X′,V′)=L(X,V). By The Joint Law of Two Classes of Square-Integrable Random Vectors: Marginals, Cost, and Invariance Under Shifts by a Vector Field (iii), L(X′)=ν, so the shifts at X′ are formed with the same E(ν) and Σ(ν) as at X, and, by the same clause with η=Σ(ν) and t=δ, respectively t=−δ, L(X,V+δΣ(ν)∘X)=L(X′,V′+δΣ(ν)∘X′) and L(X,V−δΣ(ν)∘X)=L(X′,V′−δΣ(ν)∘X′); hence Fδ−(X,r,V,X)=Fδ−(X′,r,V′,X) and Fδ+(X,r,V,X)=Fδ+(X′,r,V′,X) by the law-invariance of F.
Citations
Loading…
Dependencies
Loading…
Related
0 relations
Curated associations between results. These are editable and subjective — they do not replace the dependency graph, which is derived from the references in the text.