• Aucun résultat trouvé

A Type Theory for Linear Partial Differential Equations

N/A
N/A
Protected

Academic year: 2022

Partager "A Type Theory for Linear Partial Differential Equations"

Copied!
1
0
0

Texte intégral

(1)A Type Theory For Partial Differential Equations Marie Kerjean [email protected] IRIF, Université Paris Diderot, Paris. The Curry-Howard-Lambek correspondance Cartesian closed categories Functions f : E → F. Scientific context :. Minimal Logic Implication E ⇒ F Classical Logic ((E ⇒ ⊥) ⇒ ⊥) ' E. + ∗-autonomous categories Linear functions f ∈ L(E, F ) Reflexive vector spaces E 00 ' E. Classical Linear Logic Linear Implication E ( F [2] ((E ( ⊥) ( ⊥) ' E. + differentiation Smooth functions f ∈ C ∞(E, F ). Differential Linear Logic Linear approximation to non-linear proofs [3]. Simply Typed λ-calculus Programs acting on contexts : µαE⇒⊥. < tE |αE⇒⊥ > [1]. Differential λ-calculus. D-DiLL Resolution of Linear Partial Differential Equations [4] (LPDE). Nuclear spaces, Distributions, and Differential operators. This work :. Simply Typed λ-calculus Programs λxE .tF. Conjecture : A theorical resolution of LPDE via context-program equivalence. Denotational models of Differential Linear Logic. Differential Linear Logic. Formulas must be interpreted by reflexive vector spaces,. Linear Logic encodes non-linearity into an exponential :. E ' E 00. !E ( F ' E ⇒ F. and functions by smooth functions. And deduces curryfication for non-linear programs from the curryfication for linear one : f : C ∞(E, F ). !E⊗!F '!(E × F ).. We use the theory of topological vector spaces : Nuclear spaces 0 ( ) Fréchet DF E0 `. Rn. Differential Linear Logic allows to compute a Linear function from a non linear one : its differentiation at 0.. f ∈ C ∞(R, R). E ⊗π. D(E ⇒ F )(0) = E ( F D(f )(0). ( )0 ⊗π = ⊗ε. Fig. 2: A polarized model of MLL Fig. 1: A classification. D-DiLL The exponential rules of DiLL are :. Smooth exponentials : Distributions A typical Nuclear Frćhet space is the space of smooth functions on Rn : ∞. n. C (R , R).. ` Γ, A ` Γ, ?A d. `Γ w ` Γ, ?A. ` Γ, !A, !A c̄ ` Γ, !A. ` Γ w̄ ` Γ, !A. ` Γ, ?A, ?A c ` Γ, ?A ` Γ, A ¯ ` Γ, !A d. A typical Nuclear DF spaces is the space of distributions with compact support : C ∞(Rn, R)0 := {φ : f ∈ C ∞(Rn, R) 7→ φ(f ) ∈ R}.. The exponentials rules of D-DiLL are :. Schwartz’ Kernel Theorem : ˆ ∞(F, R)0 ' C ∞(E × F, R)0 C ∞(E, R)0⊗C !Rn = C ∞(Rn, R)0.. ` Γ wD ` Γ, ?D A ` w̄D ` !D A. ` Γ, ?A, ?D A c ` Γ, ?D A D ` Γ, !A ` ∆, !D A c̄ D ` Γ, ∆, !D A. ` Γ, ?D A ` Γ, ?A dD ` Γ, !D A ¯ ` Γ, !A d. Differential Operator on Distributions D(g)(x) =. ∂ αg |α|≤n aα (x) ∂xα .. P. !D E := (D(C ∞(E))0. Short term goals :. Theorems :. • An indexed version of D-Dill with subtyping, with an understanding at the same time of all LPDO with constant coefficients, such that c̄(!∂ i )(!∂ j ) =!∂ i+∂ j .. • If D(g) is the differentiation at 0, !D E = E 00 ' E and DiLL = sums of D − DiLL.. • It will follows from the previous system that, syntactically, !D Rn = ⊥ if and only if, semantically, the LPDE does not have a solution.. • When the coefficients aα are constant, we have a model of D-DiLL.. References 1. A formulae-as-types notion of control. Timothy G. Griffin. POPL 1990 2. Linear Logic. Jean-Yves Girard, Theoretical Computer Science , 1987.. 3. Differential interaction nets. Thomas Ehrhard, Laurent Regnier, Th. Comp. Sci., 2006. 4. A Logical Account for Linear Partial Differential Equations, K. , 2018 preprint LATEX TikZposter.

(2)

Références

Documents relatifs

Cycle cellulaire en phase végétative : division symétrique, génération de deux cellules identiques Sporulation : division asymétrique résultant en deux types cellulaires

19 Observe that uniqueness of stochastic viscosity solutions then follows from the classical theory of viscosity solutions of fully non-linear second-order partial

Also, the notion of linear rough differential equations leads to consider multiplicative functionals with values in Banach algebras more general than tensor algebras and to

We describe an explicit procedure for constructing P based on the prolongation theory of the first part of this paper which involves finitely many steps. These results imply

Our approach is based on the fact that every Frechet space F is a projective limit of Banach spaces and on the replacement of the (non-Frechet) space C (F) by an

One of the main reasons to study regularized semigroups is their flexibility in applications to evolution equations (see e.g.. We will demonstrate this in Section

(For the general theory, see the book by Godement, “Topologie alg´ebrique et th´eorie des faisceaux; the book of Ahlfors uses a less flexible presentation, resting on

d) Harmonic integrals on strongly pseudoconvex manifolds, II, Annals of Math. MORREY, A variational method in the theory of harmonic integrals, II, Amer. Journal of Math.,