Searcharxiv⌕ Search

arXiv subjects

Erhan Bayraktar

Publications and source records attributed to Erhan Bayraktar.

At least 19 recordsLinked to original sources

Adapted transport inequalities in continuous time

Given a strong solution $X$ to a stochastic differential equation (SDE), we establish sufficient conditions under which $\text{Law}(X)$ satisfies an adapted quadratic transport cost inequality (AQTCI) with respect to the semimartingale adapted Wasserstein distance. When the drift and diffusion coefficients are Markovian, we further show that the AQTCI tensorizes in a dimension-free manner. As a corollary, we show that various stochastic control problems, e.g. optimal stopping and mean-variance portfolio selection, are stable under bounded, measurable drift perturbations. Additionally, we derive concentration inequalities (with respect to certain dynamic risk measures) for stochastic integrals $\int H_s dX_s$ and deduce appropriate bounds on their tail behavior. Furthermore, using the functional Itô formula, we show that the AQTCI for $\text{Law}(X)$ allows one to deduce classical transport inequalities for $\text{Law}(F(X))$ even when $F$ is not Lipschitz on path space. Finally, we give sufficient conditions for $\text{Law}(X)$ to satisfy AQTCIs when $X$ is a solution to a scalar fractional SDE or stochastic Volterra equation.

math.PR↗

Primal--Dual Kullback--Leibler Mirror-Prox for Dynamic Mean Field Games

We compute finite-horizon dynamic mean field equilibria using a primal--dual Kullback--Leibler Mirror-Prox algorithm. Each iteration updates occupation measures by exponential tilting and computes the dual correction by solving a linear PDE, without solving a nonlinear Hamilton--Jacobi--Bellman equation or a best-response control problem. Under assumptions and Lasry--Lions monotonicity, we prove an $O(K^{-1})$ ergodic Minty bound after $K$ iterations and an $O(K^{-1/2})$ approximate-equilibrium bound for the corrected average. We obtain convergence rates for the population flow under strong monotonicity, and for smooth diffusions with uniformly convex control costs under ordinary monotonicity. We also prove convergence of continuous-time Markov chain approximations to diffusion equilibria and, under additional smoothness and convexity assumptions, bound the error in terms of the spatial mesh size and the number of iterations. Galerkin approximations preserve the $O(K^{-1})$ estimates under suitable bounds on the correction errors. Numerical experiments illustrate the method and its convergence.

math.OC↗

Price-impact Martingale optimal transport

Martingale optimal transport (MOT) provides model-independent bounds for option prices calibrated to vanilla options, assuming frictionless trading in the underlying. In this work, we consider the case where dynamic trading incurs quadratic temporary price impact. First, we prove a duality result for the superhedging price. In particular, the primal of the duality relaxes the strict martingale constraint by penalizing the deviation from a specific martingale. Second, for one asset and one trading period, we interpret this price-impact martingale optimal transport (PIMOT) as a weak optimal transport (WOT) problem, and provide a characterization in the spirit of Wasserstein projection. Finally, we justify that PIMOT is an interpolation between MOT and classical optimal transport (OT), and derive the convergence rate to both sides. Numerical examples with SPX option data show how the price bounds per claim depend on the order size.

q-fin.MF↗

Necessary and sufficient conditions for the martingale property of stochastic exponentials driven by multi-dimensional diffusions

We give necessary and sufficient coefficient conditions for stochastic exponentials driven by multidimensional diffusions to be true martingales. The classical reduction to non-exit of the diffusion with Girsanov-shifted drift separates the analysis into smooth boundary points, infinity, and nonsmooth boundary pieces. On smooth domains, under regularity and non-explosion hypotheses, martingality is equivalent to nonpositivity of the shifted Fichera drift, provided an explicit local condition holds where that drift vanishes and the stochastic integrand is unbounded. At infinity, Feller's test gives exact criteria when a scalar function of the diffusion reduces to a one-dimensional diffusion up to a time change with rate comparable to a positive function of the scalar coordinate. General radial and projection tests supply sufficient conditions. Applications give explicit necessary and sufficient conditions for inverse-CIR and $4/2$ models with correlated factors and singular stochastic integrands, including critical equality cases. The coupled non-affine examples go beyond direct scalar Feller tests and affine Riccati criteria, and exhibit boundary exits not decided by tests at infinity. Further exact criteria cover logarithmic boundary singularities, weighted distances, and an absorbed factor. The martingale conditions also yield uniqueness criteria for associated valuation equations.

math.PR↗

Deep Neural Operator Learning for Probabilistic Models

We propose a deep neural-operator framework for a general class of probabilistic models. Under global Lipschitz conditions on the operator over the entire Euclidean space-and for a broad class of probabilistic models-we establish a universal approximation theorem with explicit network-size bounds for the proposed architecture. The underlying stochastic processes are required only to satisfy integrability and general tail-probability conditions. We verify these assumptions for both European and American option-pricing problems within the forward-backward SDE (FBSDE) framework, which in turn covers a broad class of operators arising from parabolic PDEs, with or without free boundaries. Finally, we present a numerical example for a basket of American options, demonstrating that the learned model produces optimal stopping boundaries for new strike prices without retraining.

cs.LG↗

Long-time behavior of first-order Hamilton-Jacobi equations in the Wasserstein space

We establish the long-time behavior of solutions $U$ to first-order Hamilton-Jacobi equations in the Wasserstein space $\mathcal{P}_2(\mathbb T^d)$. For a class of Hamiltonians with quadratic growth, including non-convex ones, we prove that there exist a unique ergodic constant $λ$ and a solution $V$ of the associated ergodic problem such that $U(t,m)-λt$ converges to $V(m)$, uniformly on $\mathcal{P}_2(\mathbb T^d)$, as $t\to+\infty$. Finally, we provide applications to mean field control problems and zero-sum differential games.

math.AP↗

Prediction with Five Experts and Geometric Stopping: A Probabilistic Construction and Analytic Verification

We present a solution of the limiting five-expert prediction problem with geometric stopping, based on stochastic calculus and partial differential equations. If $δ$ is the stopping probability, the minimax expected regret from tied initial scores is $45π^2/(512\sqrt{2δ})+o(δ^{-1/2})$ as $δ\downarrow0$. The adversarial rank control selecting the best and third-best experts maximizes the limiting Hamiltonian at every state. Following the four-expert construction of Bayraktar, Ekren, and Zhang (2020), we represent the value correction as a discounted boundary-local-time expectation for a degenerate obliquely reflected Brownian motion. The hyperbolic systems governing its boundary traces are derived and solved explicitly. To verify the nonlinear Hamilton--Jacobi--Bellman equation, we combine heat-equation minimum principles in equality directions, ODE minimum principles applied to convex combinations of multiple controls, and comparisons across controls. These arguments reduce the 48 regional control inequalities to a few lower-dimensional boundary problems, with the remaining signs established by explicit monotonicity arguments and by factorization of multivariate integral kernels that are positive polynomials in some of their variables. We prove global $C^2$ regularity across both regional interfaces and changes in rank ordering. We also show that the alternating COMB control only attains the limiting Hamiltonian maximum exactly when the two highest scores coincide and the third- and fourth-highest scores coincide.

math.OC↗

Kinetic Mean-Field Planning via Schrödinger Bridges: Existence, Penalization, and Consistency Bounds

We study a mean-field planning problem for a swarm of interacting, non-cooperative agents whose dynamics are stochastic and kinetic. The noise acts on the velocity only, and the interaction (of Cucker-Smale type for instance) enters through the drift. For a fixed flow, the problem reduces to a Schrödinger bridge problem, and we first prove the existence of an equilibrium by a Schauder fixed-point argument. Then we replace the hard terminal constraint by a penalty function and show that equilibria of the penalized problems converge, along subsequences, to an equilibrium of the original problem. Since equilibria are not unique in general, we instead quantify the consistency error, which measures how far an equilibrium of the penalized problem is from being an equilibrium of the original one. The rate is of order $n^{-1/2}$ for the relative-entropy penalty. The proof relies on the quantitative stability of entropic optimal transport and on a pointwise linear growth bound on the score of the degenerate reference dynamics, which is of independent interest.

math.PR↗

The end-of-month option in futures contracts and the Gittins index of holding the underlying

The end-of-month option allows the short in a futures contract to choose when to deliver during a window in which the delivery price is fixed. We characterize optimal delivery through a universal index obtained from the Bank--El Karoui representation theorem. In the semimartingale setting, this is the Gittins index of holding the underlying and collecting its convenience yield, and the futures price is the retirement reward for which the value of this project equals the spot price. This characterization yields price bounds and a criterion for immediate delivery. Dinkelbach's iteration reduces the price computation to a sequence of standard optimal stopping problems, with an explicit contraction bound. We apply the theory to geometric Brownian motion and the Gibson--Schwartz model with stochastic convenience yield. For geometric Brownian motion, we characterize when the futures price equals spot and derive a square-root boundary asymptotic near expiry. For the Gibson--Schwartz model, we establish monotonicity in the yield and characterize the critical yield below which the price equals spot. Numerical results illustrate the price iteration and delivery regions.

q-fin.MF↗

Target-Generated Dirichlet Problems in State-Constrained Stochastic Control

We study state-constrained stochastic control problems with a scalar surplus $R \geq 0$. Controls for which the drift and volatility of $R$ vanish at zero define a lower-dimensional Hamilton-Jacobi-Bellman equation. Starting from control-wise state-constraint viscosity inequalities, we prove that the upper and lower limits of the interior value are a subsolution and a supersolution of this boundary equation. Terminal compatibility and comparison identify their common limit. We require one-sided bounds on positive surplus drift and generator growth, together with domination by boundary generators. For compact or coercive controls, these follow from a local lower bound and continuity of the boundary control set. For a smooth stochastic target value $w$, the transformation $R = Y - w(t,X)$ flattens the viable epigraph, and, under the stated hypotheses, the boundary control value supplies the Dirichlet datum. Applications include set-valued boundary controls, unbounded drift with quadratic costs, and redundant hedging instruments with a nonconstant target boundary.

math.OC↗

Ergodicity and turnpike properties of linear-quadratic mean field control problems

We study the asymptotic behavior of solutions to linear-quadratic mean field stochastic optimal control problems. By formulating an ergodic control framework, we characterize the convergence between the finite time horizon control problem and its ergodic counterpart. Leveraging these convergence results, we establish the turnpike property for the optimal pairs, demonstrating that solutions to the finite time horizon control problem remain exponentially close to the ergodic equilibrium except near the temporal boundaries. This result reveals the intrinsic connection between long-term dynamics and their asymptotic behavior in mean field control systems.

math.OC↗

A one-sided constrained martingale transport between two uniform laws

We minimize $\mathbb E[h(Y-X)]$ over martingale couplings of $\text{Unif}[-1,1]$ and $\text{Unif}[-2,2]$ satisfying $Y\geq X-k$, where $h\in C^1([-3,3])$ has convex derivative. Feasibility holds exactly for $k\geq1$. For each such $k$, we construct a coupling that minimizes all costs in this class. For $1<k<3$, its support consists of two graphs, with $D(x)=x-k$ on $[k-2,1]$. The maps admit an explicit parametrization for $2\leq k<3$ and are determined by scalar equations with unique admissible roots for $1<k<2$. We prove optimality by a dual inequality, using an analytic estimate in the latter range. At $k=1$ the optimizer is $Y=X\pm1$ with equal probabilities; for $k\geq3$ it is the ordinary left-curtain coupling. In the unconstrained problem, left-monotonicity identifies the left-curtain coupling, which is optimal for this cost class. Under the constraint, a discrete example shows that the corresponding support condition, even together with every two-source comparison, does not imply optimality.

math.PR↗

Viscosity and minimax solutions for path-dependent Hamilton-Jacobi equations in infinite dimensions and related differential games

We establish new results for path-dependent Hamilton-Jacobi equations with nonlinear monotone, and coercive operators on Hilbert space, which were initially studied in Bayraktar and Keller [J. Funct. Anal., 275 (8) (2018), pp. 2096-2161]. Under more general assumptions than in the cited paper (and more general than in the finite-dimensional case as well), we prove the uniqueness of a minimax solution of a terminal-value problem for the equation under consideration and the existence of such a solution on the whole path space. We introduce a new notion of a viscosity solution for this problem and show the equivalence of this notion to the notion of a minimax solution, which implies the corresponding existence and uniqueness theorem for viscosity solutions. In addition, we obtain a stability result for viscosity solutions using the half-relaxed limits method. As applications, we prove two theorems on the existence and characterization of value of a zero-sum differential game for a time-delay (path-dependent) evolution equation. The first theorem pertains to the case of non-anticipative (Elliott-Kalton) strategies and is related to the results on viscosity solutions, and the second theorem deals with the case of feedback (Krasovskii-Subbotin) strategies and is based on the results on minimax solutions.

math.AP↗

Convergence of implicit schemes for Hamilton-Jacobi-Bellman quasi-variational inequalities

In [Azimzadeh, P., and P. A. Forsyth. "Weakly chained matrices, policy iteration, and impulse control." SIAM J. Num. Anal. 54.3 (2016): 1341-1364], we outlined the theory and implementation of computational methods for implicit schemes for Hamilton-Jacobi-Bellman quasi-variational inequalities (HJBQVIs). No convergence proofs were given therein. This work closes the gap by giving rigorous proofs of convergence. We do so by introducing the notion of nonlocal consistency and appealing to a Barles-Souganidis type analysis. Our results rely only on a well-known comparison principle and are independent of the specific form of the intervention operator.

math.NA↗

High order Bellman equations and weakly chained diagonally dominant tensors

We introduce high order Bellman equations, extending classical Bellman equations to the tensor setting. We introduce weakly chained diagonally dominant (w.c.d.d.) tensors and show that a sufficient condition for the existence and uniqueness of a positive solution to a high order Bellman equation is that the tensors appearing in the equation are w.c.d.d. M-tensors. In this case, we give a policy iteration algorithm to compute this solution. We also prove that a weakly diagonally dominant Z-tensor with nonnegative diagonals is a strong M-tensor if and only if it is w.c.d.d. This last point is analogous to a corresponding result in the matrix setting and tightens a result from [L. Zhang, L. Qi, and G. Zhou. "M-tensors and some applications." SIAM Journal on Matrix Analysis and Applications (2014)]. We apply our results to obtain a provably convergent numerical scheme for an optimal control problem using an "optimize then discretize" approach which outperforms (in both computation time and accuracy) a classical "discretize then optimize" approach. To the best of our knowledge, a link between M-tensors and optimal control has not been previously established.

math.RA↗

Mean-Field PhiBE: Continuous-Time Mean-Field Reinforcement Learning from Discrete-Time Data

This paper develops a model-free framework for continuous-time mean-field control when the population evolves according to unknown controlled McKean--Vlasov dynamics and only discrete-time transition data are available. Model-based mean-field control requires the continuous-time drift and diffusion coefficients, which are not directly observed from fixed-step transitions, while a direct reduction to a discrete-time Bellman equation loses the continuous-time generator structure. To bridge these two viewpoints, we introduce a Mean-Field-PhiBE (MF-PhiBE), which incorporates discrete-time transition information into a continuous-time PDE on the Wasserstein space. The MF-PhiBE replaces the unknown infinitesimal drift and covariance in the policy-evaluation equation by one-step estimators computed from data, while preserving the generator structure of the McKean-Vlasov HJB equation. We also derive a policy-gradient theorem for entropy-regularized randomized feedback policies, expressing the actor direction through an action-wise infinitesimal advantage and the score of the policy. Combining these two ingredients yields a model-free actor-critic method. We prove a first-order consistency estimate showing that the value induced by an optimal MF-PhiBE policy approximates the optimal continuous-time value as the observation time step vanishes. For entropy-regularized LQR, we establish first-order policy convergence and second-order value convergence; under suitable conditions, the population-averaged feedback means coincide exactly. Numerical experiments on an LQR benchmark and a crowd-aversion problem illustrate the proposed framework.

math.OC↗

Convex order and preservation of convexity for Bayesian posterior updates

We study how the response of a Bayesian posterior statistic to future observations changes as information accumulates. For a non-decreasing function $T$, define $Π_n^T=\E[T(Θ)\vert \mathcal F_n]$, where $Θ$ has an arbitrary prior and the observations come from a one-parameter exponential family. Conditioning on the same current value of $Π^T$, we show that the posterior statistic after additional observations is larger in convex order when the current posterior is based on fewer observations. We also prove preservation of convexity: the expected value of a convex function of the future posterior statistic is convex in the current posterior statistic. Together, these two properties provide structural tools for establishing time-monotonicity results in dynamic Bayesian decision and optimal stopping problems. If the exponential family contains an infinitely divisible distribution, the results extend to a continuous-time observation model through a family of Lévy processes.

math.ST↗

Mean-field optimal stopping with endogenous quantile cutoffs

We study a mean-field optimal stopping problem with an endogenous population-level shutdown. All remaining agents stop when the survival mass falls below a prescribed threshold. We recast the discontinuous objective as the singular, nonconvex constraint that the survival mass lie in $\{0\}\cup[α,1]$. We prove the equivalence of strong and weak values via an approximation and the existence of an optimal rule via compactness and penalization. We also prove a dynamic programming principle. The value is continuous away from the critical boundary but may be discontinuous at the boundary itself. Under strict initial feasibility, finite-population values converge to the mean-field value. In the same regime, the laws of near-optimal empirical measures are tight and every mean-field optimizer admits a recovery sequence. At the threshold, however, finite-population convergence may fail.

math.OC↗