EN FR
EN FR


Section: New Results

Self-regulated processes

Participants : Jacques Lévy Véhel, Anne Philippe, Caroline Robet

We wish to construct various instances of processes Z such that, at each point t, almost surely, the pointwise Hölder exponent of Z at t, denoted αZ(t), verifies

α Z ( t ) = g ( Z ( t ) )

where g∈𝒞1(ℝ,[a,b]) is a deterministic function. Then, we would estimate the function g which control the regularity.

The pointwise Hölder exponent at t of a function or a process f:ℝ→ℝ, which is 𝒞1 nowhere, is the real αf(t) such that :

α f ( t ) = sup { β , lim sup h → 0 ∣ f ( t + h ) - f ( t ) ∣ ∣ h ∣ β = 0 }

We worked first on pathwise integrals :

Theorem 1 Let g∈𝒞1(ℝ,[a,b]), 0<a<b<1. Provided ∥g'∥∞ is small enough, there exists a unique continuous process Z verifying almost surely on [0,T]

Z t = ∫ 0 t ( t - u ) g ( Z u ) - 1 W u d u

where W is an almost surely continuous process.

A random condition (∥g'∥∞∥W(ω)∥∞C(a,T)<1) appears in the application of Banach fixed point theorem (in (𝒞0([0,T];ℝ),∥.∥∞)). It implies that it is possible to have existence et uniqueness only on [0,t'], t'<T. We simulated pathwise integrals and showed some cases without uniqueness. We studied some easier processes in order to find the regularity of Z.

Theorem 2 Let h∈]0,1[ and U defined on [0,T] by

U t = ∫ 0 t ( t - u ) h - 1 W u d u

Then ∀t∈[0,T], αU(t)≥h.

Theorem 3 Let g∈𝒞1(ℝ,[a,b]), 0<a<b<1. Provided ∥g'∥∞ is small enough, there exists a unique continuous process Y verifying almost surely on [0,T]

Y t = ∫ 0 t ( t - u ) g ( Y t ) - 1 W u d u

where W is an almost surely continuous process. Furthermore, ∀t∈[0,T], αY(t)≥g(Yt)

Then, we adapted the multifractional Brownian Motion [50], [31] (which a representation is Bt=∫0tKH(t)(t,u)W(du), W Brownian Motion et H∈𝒞1) to construct the modified multifractional Brownian Motion : Zt=∫0tKH(u)(t,u)W(du). We expect obtain a self-regulated process Yt=∫0tKg(Yu)(t,u)dW(u).

Theorem 4 Let g∈𝒞1(ℝ,[a,b]), 0<a<b<1. Provided ∥g'∥∞ is small enough, there exists a unique continuous adapted process Y include in 𝒞0([0,T];L2(Ω)) verifying almost surely on [0,T]

Y t = ∫ 0 t K g ( Y u ) ( t , u ) d W ( u )

where W is the Brownian motion.