Searcharxiv⌕ Search

arXiv subjects

Alexandre Autran

Publications and source records attributed to Alexandre Autran.

4 recordsLinked to original sources

Quadrature-Induced Local Minima in Structure-Preserving Lotka-Volterra Control

Harvest quadrature can change a discrete optimal control problem even when the state scheme stays fixed. We study this effect for a positivity-preserving Lotka--Volterra scheme. The difference appears in a model with harvesting alone: rewarding the scheme's removal flux gives a strictly convex reduced cost, whereas using prey density at the start of each step overestimates harvest and can create negative curvature and a strict local minimum that is not global. Rational root isolation certifies the two-step global minimizers. With diffusion in one space dimension, nonnegative bounded $H^1$ initial data and positive diffusivities, the costs, weighted gradients and Hessians differ by $O(τ)$ uniformly in the spatial mesh, where $τ$ is the time step. A uniform positive lower bound on the reduced Hessian gives the same order for the distance between nearby controls. A large enough effort penalty gives unique global discrete minimizers. We also prove positivity, exact population balance and a common variational limit, and derive exact discrete adjoints and Hessian actions. Numerical tests trace changes in curvature and stationary controls across parameters and grids, with costs evaluated on a common reference using adaptive time stepping. The construction extends to weighted production--destruction systems, giving exact harvest accounting and smooth derivatives on fixed grids.

math.NA↗

Singular Tracking of Fractional Brownian Motion under Proportional Costs

We study the tracking of fractional Brownian motion by adapted controls of finite variation, with quadratic loss and proportional adjustment costs. Let $B^H$ be standard fractional Brownian motion with Hurst parameter $H\in(0,1)$. We write $V_λ(T;a,b,σ)$ for the optimal expected cost of tracking $σB^H$ over a horizon $T>0$, where $σ>0$ is the target amplitude, $a>0$ weights the quadratic loss, and $λb$ is the cost per unit of adjustment, with $λ,b>0$. We prove \[V_λ(T;a,b,σ)\underset{λ\to0}{\sim}κ_HT a^{(1-H)/(1+H)}(λb)^{2H/(1+H)}σ^{2/(1+H)}. \] Here $κ_H>0$ is the long-run optimal cost per unit time in the normalized problem $a=b=σ=λ=1$. Observing the entire past does not change $κ_H$. For the normalized problem with full past information, a horizon $L>0$ and a terminal adjustment to the target, the difference between its value and $κ_H L$ is bounded uniformly in $L$. This estimate also yields continuity of $κ_H$ in $H$. In the Brownian case, $κ_{1/2}=\frac12(\frac32)^{2/3}$. We extend the asymptotic formula to nonstationary targets under a coupling condition on local paths and their observation filtrations. We verify this condition for Gaussian Volterra targets under Brownian observation. All asymptotically optimal controls have the same leading-order split between tracking and adjustment costs. We also give primal and dual bounds on $κ_H$. For $H<1/2$, we construct approximations based on exponential sums with positive weights and quantify their error as the horizon grows.

math.OC↗

Conditional Fluctuations of Mean-Field Systems with Stable Common Noise

We study conditional fluctuations of smooth mean-field systems on a torus with symmetric stable common jumps of index $0<α<1$. A positive-strip estimate makes empirical Poisson acceptance errors negligible at the sampling scale. Given the full Poisson master, the conditional laws converge in probability to a centered Gaussian kernel in a negative Sobolev path space. In dimension one, quantile coupling preserves this limit for microscopic marks with relative tail remainder $O(x^{-ρ})$, where $ρ>\tfracα{2}$. A second-order tail expansion identifies the response when this condition fails: it shifts the conditional Gaussian mean at criticality and, when nonzero, dominates sampling below criticality. We also identify the transition from a stable to a Brownian rank error. In a translation model, rank and clock couplings give the same microscopic laws but different path fluctuations. Rank coupling gives a functional Gaussian limit. Clock coupling admits no diverging additive path normalization at the limiting center.

math.PR↗

Second-order Wasserstein response for Lévy laws

Moving an atom and splitting it symmetrically can give the same weak change in a Lévy characteristic, yet produce Wasserstein errors of different orders. We explain this through first- and second-order expansions of the distribution function for infinitely divisible laws with a Gaussian factor. The expansions are uniform over Lipschitz tests and include distributional directions from moving atoms. When the first response is nonzero, analyticity gives the signed second coefficient of the Wasserstein--1 distance. When it vanishes, a finite second displacement moment gives a quadratic response, with a positive coefficient whenever the dispersion is nonzero. For local balanced remeshing, we prove matching bounds in terms of grid-alignment variance and construct a linear program with a controlled optimization gap. We also determine the sharp transition as Gaussian smoothing vanishes. The appendices give proofs, metric speed and length, and multivariate and state-dependent extensions.

math.PR↗