Function Approximation Research Articles

A Critical Evaluation of the Hybrid KS DFT Functionals Based on the KS Exchange-Correlation Potentials.

We have developed a critical methodology for the evaluation of the quality of hybrid exchange-correlation (XC) density functional approximations (DFAs) based on very fundamental quantities, i.e., Kohn-Sham (KS) XC potentials, self-consistent electron densities, first ionization potentials (IPs), and total energies. Since the XC potentials, the primary objects in the current study, are not directly accessible for the hybrids, we calculate them by inverting the KS electron densities. Utilizing this methodology, we tested 155 hybrid DFAs available in the LIBXC library using FCI and CCSD(T) methods as a reference. We have found that a group of functionals produces very decent XC potentials, mainly those with a large mixture of Hartree-Fock exchange. Moreover, the value of IP strongly depends on the XC potential quality. On the other hand, we show that the XC energy is dominated by functional-driven error, which in some cases leads to substantial errors in electronic densities. The study shows new directions for constructing more accurate XC functionals within the KS-DFT framework.

Learning on manifolds without manifold learning

Function approximation based on data drawn randomly from an unknown distribution is an important problem in machine learning. The manifold hypothesis assumes that the data is sampled from an unknown submanifold of a high dimensional Euclidean space. A great deal of research deals with obtaining information about this manifold, such as the eigendecomposition of the Laplace–Beltrami operator or coordinate charts, and using this information for function approximation. This two-step approach implies some extra errors in the approximation stemming from estimating the basic quantities of the data manifold in addition to the errors inherent in function approximation. In this paper, we project the unknown manifold as a submanifold of an ambient hypersphere and study the question of constructing a one-shot approximation using a specially designed sequence of localized spherical polynomial kernels on the hypersphere. Our approach does not require preprocessing of the data to obtain information about the manifold other than its dimension. We give optimal rates of approximation for relatively “rough” functions.

Numerical differentiation of the piecewise smooth function by using Fourier extension method

Numerical differentiation of the piecewise smooth function is considered in this paper. To avoid the large error of numerical differentiation that may occur near potential non-smooth points, we identify the discontinuity points of the first or second derivative of the function. Then we divide the domain of the function into several sub-domains. For each sub-domain, the approximation is constructed by Fourier extension, and the global approximation of the piecewise smooth function is formed by superposition to improve accuracy. Some numerical experiments are conducted to further verify the efficacy of the method.

Intractability results for integration in tensor product spaces

We prove lower bounds on the worst-case error of numerical integration in tensor product spaces. The information complexity is the minimal number N of function evaluations that is necessary such that the N-th minimal error is less than a factor ε times the initial error, i.e., the error for N=0, where ε belongs to (0,1). We are interested to which extent the information complexity depends on the number d of variables of the integrands. If the information complexity grows exponentially fast in d, then the integration problem is said to suffer from the curse of dimensionality.Under the assumption of the existence of a worst-case function for the uni-variate problem, we present two methods for providing lower bounds on the information complexity. The first method is based on a suitable decomposition of the worst-case function and can be seen as a generalization of the method of decomposable reproducing kernels. The second method, although only applicable for positive quadrature rules, does not require a suitable decomposition of the worst-case function. Rather, it is based on a spline approximation of the worst-case function and can be used for analytic functions. Several applications of both methods are presented.

Whittle Index-Based Q-Learning for Wireless Edge Caching With Linear Function Approximation

Whittle Index-Based Q-Learning for Wireless Edge Caching With Linear Function Approximation

Adaptive sliding mode control of petrochemical flare combustion process based on radial basis function network

Steam-assisted combustion elevated flares are currently the most widely used type of petrochemical flares. Due to the complex and variable composition of the waste gas they handle, the combustion environment is severely affected by meteorological conditions. Key process parameters such as intake composition, flow rate, and real-time data of post-combustion residues are difficult to measure or exhibit lag in data availability. As a result, the control methods for these flares are limited, leading to poor control effectiveness. To address this issue, this paper proposes an adaptive sliding mode control method based on the Radial Basis Function (RBF) network. Firstly, the operational characteristics of the petrochemical flare combustion process are analyzed, and a control model for the combustion process is established based on carbon dioxide detection. Secondly, an RBF neural network-based unknown function approximator is designed to identify the nonlinear part of the actual operating system. Finally, by combining the control model of the petrochemical flare combustion and designing the RBF sliding mode controller with its adaptive control law, fast and stable control of the flare combustion state is achieved. Simulation results demonstrate that the designed control strategy can achieve tracking control of the petrochemical flare combustion state, and the adaptive law also accomplishes system identification.

Neutral donors confined in semiconductor coupled quantum dot-rings: Position-dependent properties and optical transparency phenomenon

Electronic properties of a neutral donor confined in a GaAs coupled quantum dot-ring covered by a Al0.3Ga0.7As matrix were calculated using the finite element method under the effective mass and the envelope function approximations. The proposed model is set up to fit a realistic coupled quantum dot-ring geometry revealed by atomic force microscopy images. The results show that the energy levels and the transition energies in the presence of an electric field strongly depend on the donor center’s angular position. Furthermore, the total optical absorption coefficient is calculated within the two-level approximation and the matrix density formalism. The absorption spectrum shows that the system can be tuned between 5 and 30meV. Also, an optical transparency effect for different configurations characterized by specific donor center’s angular positions and electric field values is seen. Finally, a novel redshift is observed when the sample temperature increases.

Dynamic pickup-and-delivery for collaborative platforms with time-dependent travel and crowdshipping

We study a pickup-and-delivery problem that arises when customers randomly submit requests over the course of a day from a choice of vendors on a collaborative e-commerce portal. Based on the attributes of a customer request, a dispatcher dynamically schedules the delivery service on either a dedicated vehicle or a crowdshipper, both of whom experience time-dependent travel times. While dedicated vehicles are available throughout the day, the availability of crowdshippers is unknown a priori and they appear randomly for only portions of the day. With an objective of minimizing the sum of routing costs, piece-rate crowdshipper payments, and lateness charges, we model the uncertainty in request arrivals and crowdshipper appearances as a Markov decision process. To determine an action at each decision epoch, we employ a heuristic that partially destroys the existing routes and repairs them under the guidance of a parameterized cost function approximation that accounts for the remaining temporal capacity of delivery vehicles. We benchmark our real-time heuristic with an adaptive large neighborhood search and demonstrate the effectiveness of our method with several performance metrics. In addition, we conduct computational experiments to demonstrate the impact of inserting wait time in the route scheduling and the benefit of explicitly modeling time-dependent travel times. Through our computational testing, we also investigate the potential of demand management mechanisms that facilitate many-to-one request bundles or one-to-many request bundles to reduce the cost to service requests.

An Overview Of Deep Deterministic Policy Gradient Algorithm And Applications

This research summarizes the major developments of the Deep Deterministic Policy Gradient (DDPG) in reinforcement learning Motivated by the ideas, of Deep Q-networks, DDPG has proven capable of confronting more complex problems involving continuous action spaces. The core of DDPG lies in its actor-critic architecture, which enables the learning of highly competitive policies. By leveraging neural network function approximations, it can efficiently operate in large state and action spaces. DDPG has found practical applications across various real-world domains. However, like many model-free reinforcement learning methods, DDPG still faces the challenge of requiring a large number of training steps.

ADRP-DQL: An adaptive distributed routing protocol for underwater acoustic sensor networks using deep Q-learning

Underwater Wireless Sensor Networks (UWSNs) face unique constraints due to their unstructured and dynamic underwater environment. Data gathering from these networks is crucial as energy resources are limited. In this regard, efficient routing protocols are needed to optimize energy consumption, increase the network lifetime, and enhance data delivery in these networks. In this work, we develop an Adaptive Distributed Routing Protocol for UWSNs using Deep Q-Learning (ADRP-DQL). This protocol employs the ability of reinforcement learning to dynamically learn the best routing decisions based on the network’s state and action-value estimates. It allows nodes to make intelligent routing decisions, considering residual energy, depth and node degree. A Deep Q-Network (DQN) is employed as the function approximator to estimate action values and choose the optimal routing decisions. The DQN is trained using off-policy and on-policy strategies and the neural network model. Simulation results demonstrate that ADRP-DQL performs well regarding energy efficiency (EE), data delivery ratio, and network lifetime. The results highlight the proposed protocol’s effectiveness and adaptability to UWSNs. The ADRP-DQL protocol contributes to intelligent routing for UWSNs, offering a promising approach to enhance performance and optimize energy utilization in these demanding environments.

Low-Complexity Tracking Control of Unknown Strict-Feedback Systems With Quantitative Performance Guarantees.

This article is concerned with the prescribed performance tracking control problem for the strict-feedback systems with unknown nonlinearities and unmatched disturbances. The challenge lies in the realization of a complete performance specification for trajectory tracking in the sense of quantitatively regulating the peak value, overshoot, settling time, and accuracy while ensuring that the initial condition holds naturally. To this end, an error transformation, equipped with a shifting function, is introduced and incorporated with a new-type barrier function. Then, a class of performance functions is exploited to quantify the settling times and steady-state bounds of the intermediate errors. Moreover, to improve the flexibility of formulating performance specifications for the tracking error, a pair of asymmetric performance boundaries are further designed. With their combination, a novel robust prescribed performance control (PPC) approach is proposed in this article. It not only achieves the quantitative performance guarantees but also preserves the unique simplicity of PPC, evading the needs for function approximation, parameter identification, disturbance estimation, derivative calculation, or command filtering. The above theoretical findings are confirmed via three simulation studies.

Randomized semi-quantum matrix processing

We present a hybrid quantum-classical framework for simulating generic matrix functions more amenable to early fault-tolerant quantum hardware than standard quantum singular-value transformations. The method is based on randomization over the Chebyshev approximation of the target function while keeping the matrix oracle quantum, and is assisted by a variant of the Hadamard test that removes the need for post-selection. The resulting statistical overhead is similar to the fully quantum case and does not incur any circuit depth degradation. On the contrary, the average circuit depth is shown to get smaller, yielding equivalent reductions in noise sensitivity, as explicitly shown for depolarizing noise and coherent errors. We apply our technique to partition-function estimation, linear system solvers, and ground-state energy estimation. For these cases, we prove advantages on average depths, including quadratic speed-ups on costly parameters and even the removal of the approximation-error dependence.

Best one sided approximation of unbounded functions trigonometrical polynomials for one dimensional weighted-space

The aim of studying this research is to find the best one sided trigonometrical approximation of unbounded function of one variable in weighted space. Moreover we estimated the degree of the best one-sided trigonometrical approximation by averaged modulus of smoothness in same space.

Найкраще наближення в інтегральній метриці лінійної комбінації періодичних функцій різної парності

In many areas of mathematics, extremal problems often arise related to approximation characteristics of both the approximating functions and the properties of the elements being approximated. For example, in the case of polynomials, the problem can involve the number of points where the values of the function coincide with the values of the polynomial used to replace the function on the studied interval. In practice, the problem of approximating a function from a given set R is reduced to replacing it, according to a defined algorithm, with a function (polynomial) of a fixed degree that is close to it in a certain sense. In uniform and integral metrics, the problem of finding the exact values of the best approximations for classes of r-times differentiable functions, where r is a natural number, has been explored in the works of J. Favard [1], N. I. Akhiezer, M. G. Krein [2], B. Nadi, S. M. Nikolsky [3], V. K. Dzyadyk [4], S. B. Stechkin, Sun Yun-shen, and others. The final results concerning the solution of the best approximation problem on Weyl-Nagy classes for arbitrary values of the parameters defining these classes belong to the Ukrainian mathematician V. K. Dzyadyk [4]. In his works on function classes generated by the well-known Bernoulli kernels, it was established that the number of coincidence points between the kernel and the approximating polynomial of degree n − 1, including their multiplicities, does not exceed 2n, which allowed for obtaining final results. The work [5] presents cases of such linear combinations of even or odd kernels for which the number of uniformly distributed interpolation points equals 2n + 2 for a polynomial of degree n − 1 that deviates the least in the metric of the L – space from the studied linear combination. The idea of studying composite kernels expressed as a linear combination of component terms belongs to O. I. Stepanyets [6], and it was implemented in problems of joint approximation of functions and their derivatives. In the 1980s and 1990s, O. I. Stepanets developed a new approach to classifying periodic functions, which allowed for a fine classification of extremely broad sets of periodic functions. The results obtained for these classes are, on one hand, of a general nature, and on the other, they provide a whole series of new, previously unknown results that could not be achieved with previously known classes. Following the approaches to function classification, we can consider a linear combination of function classes as a certain single class of a more complex nature. Then, the problem of finding the exact values of the upper bounds of the best approximations reduces to the problem of the best approximation for this composite class, which corresponds to convolutions with the generating composite kernel. In this work, we investigate linear combinations of three continuous 2π-periodic kernels of different parity, and it is established that there exist such composite kernels that their trigonometric polynomial of the best approximation of order n − 1 in the integral metric interpolates the kernel at 2n + 2 uniformly distributed points

LinFa-Q: Accurate Q-learning with linear function approximation

Although Q-learning has achieved remarkable success in some practical cases, it often suffers from the overestimation problem in stochastic environments, which is commonly viewed as a shortcoming of Q-learning. Overestimated values are introduced by estimations on the next state Q value, which is well-known as the maximization bias. In this paper, we propose a more accurate method for estimating the Q value by the Q value decomposition and re-evaluation with similar samples based on linear function approximation. Specifically, we reform the parameterized incremental update formula of Q-learning and also demonstrate that the new formula is equivalent to the original one. Moreover, we propose a new parameterized incremental update formula of Q-learning to address the overestimation problem and present the more accurate computing method, which can be used in problems with continuous state spaces and stochastic environments. Experimentally, when compared with Doubly Bounded Q-learning and other Q-learning based methods, the new algorithm has more than 31% improvement of performance in Mountain Car and Cart Pole. Furthermore, the algorithm is robust to the learning rate and its memory capacity. Finally, the practical applicability of our algorithm is discussed through an analysis of time consumption.

Electron Localization Function for Noncollinear Spins.

Understanding of bonding is key to modeling materials and predicting properties thereof. A widely adopted indicator of bonds and atomic shells is the electron localization function (ELF). The building blocks of the ELF are also used in the construction of modern density functional approximations. Here, we demonstrate that the ELF breaks down when applied beyond regular nonrelativistic quantum states. We show that for tackling general noncollinear open-shell solutions, it is essential to address both the U(1) gauge invariance, i.e., invariance under a multiplication by a position dependent phase factor, and SU(2) gauge invariance, i.e., invariance under local spin rotations, conjointly. Remarkably, we find that the extended ELF also improves the description of paradigmatic collinear states.

Chebyshev–Padé approximants for multivalued functions

The paper discusses the connection between the linear Chebyshev–Padé approximants for an analytic function f f and diagonal type I Hermite–Padé polynomials for the set of functions [ 1 , f 1 , f 2 ] [1, f_1, f_2] , where the pair of functions f 1 , f 2 f_1, f_2 forms a Nikishin system. Both problems can ultimately be reduced to certain convergence problems for multipoint Padé approximants. On the other hand, the denominators of multipoint Padé approximants are non-Hermitian orthogonal polynomials with analytical weights. Thus, to study all the above problems, the general method created by Herbert Stahl can be applied. Stahl’s method is not yet sufficiently developed to obtain general results on these problems. In particular, many key convergence problems for Chebyshev–Padé approximants for functions with arbitrary configurations of branch points remain open. In this paper, we consider several important general and particular results related to this case, some already well known, and also formulate two general hypotheses in the indicated direction.

The Gonchar–Chudnovskies conjecture and a functional analogue of the Thue–Siegel–Roth theorem

This article examines the Gonchar–Chudnovskies conjecture about the limited size of blocks of diagonal Padé approximants of algebraic functions. The statement of this conjecture is a functional analogue of the famous Thue–Siegel–Roth theorem. For algebraic functions with branch points in general position, we will show the validity of this conjecture as a consequence of recent results on the uniform convergence of the continued fraction for an analytic function with branch points. We will also discuss related problems on estimating the number of “spurious” (“wandering”) poles for rational approximations (Stahl’s conjecture), and on the appearance and disappearance of defects (Froissart doublets).

Ab Initio Calculation of Coupling-Constant Averaged Exchange-Correlation Holes for Spherically Symmetric Atoms.

Accurate approximation of the exchange-correlation (XC) energy in density functional theory (DFT) calculations is essential for reliably modeling electronic systems. Many such approximations are developed from models of the XC hole; accurate reference XC holes for real electronic systems are crucial for evaluating the accuracy of these models however the availability of reliable reference data is limited to a few systems. In this study, we employ the Lieb optimization with a coupled cluster singles and doubles (CCSD) reference to construct accurate coupling-constant averaged XC holes, resolved into individual exchange and correlation components, for five spherically symmetric atoms: He, Li, Be, N, and Ne. Alongside providing a new set of reference data for the construction and evaluation of model XC holes, we compare our data against the exchange and correlation hole models of the established local density approximation (LDA) and Perdew-Burke-Ernzerhof (PBE) density functional approximations. Our analysis confirms the established rationalization for the limitations of LDA and the improvement observed with PBE in terms of the hole depth and its long-range decay, which is demonstrated in real-space for this series of spherically symmetric atoms.

Numerical analysis of small-strain elasto-plastic deformation using local Radial Basis Function approximation with Picard iteration

In this paper, we discuss a von Mises plasticity model with nonlinear isotropic hardening assuming small strains in a plane strain example of internally pressurised thick-walled cylinder subjected to different loading conditions. The elastic deformation is modelled using the Navier-Cauchy equation. In regions where the von Mises stress exceeds the yield stress, corrections are made locally through a return mapping algorithm. We present a novel method that uses a Radial Basis Function-Finite Difference (RBF-FD) approach with Picard iteration to solve the system of nonlinear equations arising from plastic deformation. This technique eliminates the need to stabilise the divergence operator and avoids special positioning of the boundary nodes, while preserving the elegance of the meshless discretisation and avoiding the introduction of new parameters that would require tuning. The results of the proposed method are compared with analytical and Finite Element Method (FEM) solutions. The results show that the proposed method achieves comparable accuracy to FEM while offering significant advantages in the treatment of complex geometries without the need for conventional meshing or special treatment of boundary nodes or differential operators.