A Radial Basis Function Method for Global Optimization

We introduce a method that aims to find the global minimum of a continuous nonconvex function on a compact subset of . It is assumed that function evaluations are expensive and that no additional information is available. Radial basis function interpolation is used to define a utility function. The maximizer of this function is the next point where the objective function is evaluated. We show that, for most types of radial basis functions that are considered in this paper, convergence can be achieved without further assumptions on the objective function. Besides, it turns out that our method is closely related to a statistical global optimization method, the P-algorithm. A general framework for both methods is presented. Finally, a few numerical examples show that on the set of Dixon-Szegö test functions our method yields favourable results in comparison to other global optimization methods.     

Radial Basis Function methods—reduced computational expense by exploiting symmetry

Radial basis function (RBF) methods are popular methods for scattered data interpolation and for solving PDEs in complexly shaped domains. RBF methods are simple to implement as they only require elementary linear algebra operations. In this work, center locations that result in matrices with a centrosymmetric structure are examined. The resulting matrix structure can be exploited to reduce computational expense and improve the accuracy of the methods while avoiding more complicated techniques such as domain decomposition.     

Deep Reinforcement Learning for Market Making in Corporate Bonds: Beating the Curse of Dimensionality

ABSTRACT

In corporate bond markets, which are mainly OTC markets, market makers play a central role by providing bid and ask prices for bonds to asset managers. Determining the optimal bid and ask quotes that a market maker should set for a given universe of bonds is a complex task. The existing models, mostly inspired by the Avellaneda-Stoikov model, describe the complex optimization problem faced by market makers: proposing bid and ask prices for making money out of the difference between them while mitigating the market risk associated with holding inventory. While most of the models only tackle one-asset market making, they can often be generalized to a multi-asset framework. However, the problem of solving the equations characterizing the optimal bid and ask quotes numerically is seldom tackled in the literature, especially in high dimension. In this paper, we propose a numerical method for approximating the optimal bid and ask quotes over a large universe of bonds in a model à la Avellaneda–Stoikov. As classical finite difference methods cannot be used in high dimension, we present a discrete-time method inspired by reinforcement learning techniques, namely, a model-based deep actor-critic algorithm.     

基于多元二次径向基神经网络的偏微分求解方法

为提高偏微分方程的计算求解精度,设计了以多元二次径向基神经网络为求解单元的偏微分计算方法,给出了多元二次径向基神经网络的具体求解结构,并以此神经网络为求解基础,给出了具体的偏微分计算步骤.通过具体的偏微分求解实例验证方法的有效性,并以3种不同设计样本数构建的多元二次径向基神经网络为计算单元,从实例求解所需的计算时间以及解的精度作对比,结果表明,采用基于多元二次径向基神经网络的偏微分方程求解方法具有求解精度高以及计算效率低等特点.     

强化学习在运筹学的应用:研究进展与展望

强化学习已经成为人工智能领域一个新的研究热点,并已成功应用于各领域,强化学习将运筹优化领域的很多问题视为序贯决策问题,建模为马尔可夫决策过程并进行求解,在求解复杂、动态、随机运筹优化问题具有较大的优势。本文主要对强化学习在运筹优化领域的应用进行综述,首先介绍了强化学习的基本原理及其应用于运筹优化领域的研究框架,然后回顾并总结了强化学习在库存控制、路径优化、装箱配载和车间作业调度等方面的研究成果,并将最新的深度强化学习以及传统方法在运筹学领域的应用研究进行了对比分析,以突出深度强化学习的优越性。最后提出几个值得进一步探讨的研究方向,期望能为强化学习在运筹优化领域的研究提供参考。     

Application of a Near-Optimal Reinforcement Learning Controller to a Robotics Problem in Manufacturing: A Hybrid Approach

Optimization theory provides a framework for determining the best decisions or actions with respect to some mathematical model of a process. This paper focuses on learning to act in a near-optimal manner through reinforcement learning for problems that either have no model or the model is too complex. One approach to solving this class of problems is via approximate dynamic programming. The application of these methods are established primarily for the case of discrete state and action spaces. In this paper we develop efficient methods of learning which act in complex systems with continuous state and action spaces. Monte-Carlo approaches are employed to estimate function values in an iterative, incremental procedure. Derivative-free line search methods are used to obtain a near-optimal action in the continuous action space for a discrete subset of the state space. This near-optimal control policy is then extended to the entire continuous state space via a fuzzy additive model. To compensate for approximation errors, a modified procedure for perturbing the generated control policy is developed. Convergence results under moderate assumptions and stopping criteria are established.     

Dynamic analysis of plane elasticity with new complex Fourier radial basis functions in the dual reciprocity boundary element method

In this paper, dual reciprocity (DR) boundary element method (BEM) is reformulated using new radial basis function (RBF) to approximate the inhomogeneous term of Navier’s differential equation (i.e., inertia term). This new RBF, which is in the form of exp(iωr), is called complex Fourier RBF hereafter. The present RBF has simultaneously collected the properties of Gaussian and real Fourier RBF reported in literature together. Consequently, this promising feature has provided more robustness and potency of the proposed method. The required kernels for displacement and traction particular solutions are derived by employing the method of variation of parameters. As some terms of these kernels are singular, a new simple smoothing trick is employed to resolve the singularity problem. Moreover, the limiting values of relevant kernels are evaluated. The validity, accuracy, and strength of the present formulation are illustrated throughout several numerical examples. The numerical results show that the proposed complex Fourier RBF represents more accurate solutions, using less degree of freedom compared to other RBFs available in the literature.     

The radial basis functions method for identifying an unknown parameter in a parabolic equation with overspecified data

Parabolic partial differential equations with overspecified data play a crucial role in applied mathematics and engineering, as they appear in various engineering models. In this work, the radial basis functions method is used for finding an unknown parameter p(t) in the inverse linear parabolic partial differential equation u_t = u_xx + p(t)u + φ, in [0,1] × (0,T], where u is unknown while the initial condition and boundary conditions are given. Also an additional condition ∫₀¹k(x)u(x,t)dx = E(t), 0 ≤ t ≤ T, for known functions E(t), k(x), is given as the integral overspecification over the spatial domain. The main approach is using the radial basis functions method. In this technique the exact solution is found without any mesh generation on the domain of the problem. We also discuss on the case that the overspecified condition is in the form ∫₀^s(t) u(x,t)dx = E(t), 0 < t ≤ T, 0 < s(t) < 1, where s and E are known functions. Some illustrative examples are presented to show efficiency of the proposed method. © 2007 Wiley Periodicals, Inc. Numer Methods Partial Differential Eq, 2007     

RBF‐ENO/WENO schemes with Lax–Wendroff type time discretizations for Hamilton–Jacobi equations

In this research, a class of radial basis functions (RBFs) ENO/WENO schemes with a Lax–Wendroff time discretization procedure, named as RENO/RWENO‐LW, for solving Hamilton–Jacobi (H–J) equations is designed. Particularly the multi‐quadratic RBFs are used. These schemes enhance the local accuracy and convergence by locally optimizing the shape parameters. Comparing with the original WENO with Lax–Wendroff time discretization schemes of Qiu for HJ equations, the new schemes provide more accurate reconstructions and sharper solution profiles near strong discontinuous derivative. Also, the RENO/RWENO‐LW schemes are easy to implement in the existing original ENO/WENO code. Extensive numerical experiments are considered to verify the capability of the new schemes.     

Regularized Lotka-Volterra Dynamical System as Continuous Proximal-Like Method in Optimization

We introduce and study a new type of dynamical system which combines the continuous gradient method with a nonlinear Lotka-Volterra (LV) type of differential system within a logarithmic-quadratic proximal scheme. We prove a global existence and viability result for the resulting trajectory which holds for a general smooth function. The asymptotic behavior of the produced trajectory is analyzed and global convergence of the trajectory to a minimizer of the convex minimization problem over the nonnegative orthant is established. The implicit discretization which is at the origin of the proposed continuous dynamical system is an interior proximal scheme for minimizing a closed proper convex function, and convergence results and properties of the resulting discrete scheme are also established. We show finally that the trajectories of the family of regularized Lotka-Volterra systems, parametrized by the positive parameter associated with the quadratic proximal term, are uniformly convergent to the solution of the classical LV-dynamical system, as the parameter associated with the proximal term approaches zero.