Stochastic Approximation Algorithms for Constrained Optimization via Simulation

We develop four algorithms for simulation-based optimization under multiple inequality constraints. Both the cost and the constraint functions are considered to be long-run averages of certain state-dependent single-stage functions. We pose the problem in the simulation optimization framework by using the Lagrange multiplier method. Two of our algorithms estimate only the gradient of the Lagrangian, while the other two estimate both the gradient and the Hessian of it. In the process, we also develop various new estimators for the gradient and Hessian. All our algorithms use two simulations each. Two of these algorithms are based on the smoothed functional (SF) technique, while the other two are based on the simultaneous perturbation stochastic approximation (SPSA) method. We prove the convergence of our algorithms and show numerical experiments on a setting involving an open Jackson network. The Newton-based SF algorithm is seen to show the best overall performance.

On the performance of strain smoothing for quadratic and enriched finite element approximations (XFEM/GFEM/PUFEM)

By using the strain smoothing technique proposed by Chen et al. (Comput. Mech. 2000; 25: 137-156) for meshless methods in the context of the finite element method (FEM), Liu et al. (Comput. Mech. 2007; 39(6): 859-877) developed the Smoothed FEM (SFEM). Although the SFEM is not yet well understood mathematically, numerical experiments point to potentially useful features of this particularly simple modification of the FEM. To date, the SFEM has only been investigated for bilinear and Wachspress approximations and is limited to linear reproducing conditions. The goal of this paper is to extend the strain smoothing to higher order elements and to investigate numerically in which condition strain smoothing is beneficial to accuracy and convergence of enriched finite element approximations. We focus on three widely used enrichment schemes, namely: (a) weak discontinuities; (b) strong discontinuities; (c) near-tip linear elastic fracture mechanics functions. The main conclusion is that strain smoothing in enriched approximation is only beneficial when the enrichment functions are polynomial (cases (a) and (b)), but that non-polynomial enrichment of type (c) lead to inferior methods compared to the standard enriched FEM (e.g. XFEM). Copyright (C) 2011 John Wiley & Sons, Ltd.

Monte-Carlo estimation of time-dependent statistical characteristics of random dynamical systems

The problem of estimating the time-dependent statistical characteristics of a random dynamical system is studied under two different settings. In the first, the system dynamics is governed by a differential equation parameterized by a random parameter, while in the second, this is governed by a differential equation with an underlying parameter sequence characterized by a continuous time Markov chain. We propose, for the first time in the literature, stochastic approximation algorithms for estimating various time-dependent process characteristics of the system. In particular, we provide efficient estimators for quantities such as the mean, variance and distribution of the process at any given time as well as the joint distribution and the autocorrelation coefficient at different times. A novel aspect of our approach is that we assume that information on the parameter model (i.e., its distribution in the first case and transition probabilities of the Markov chain in the second) is not available in either case. This is unlike most other work in the literature that assumes availability of such information. Also, most of the prior work in the literature is geared towards analyzing the steady-state system behavior of the random dynamical system while our focus is on analyzing the time-dependent statistical characteristics which are in general difficult to obtain. We prove the almost sure convergence of our stochastic approximation scheme in each case to the true value of the quantity being estimated. We provide a general class of strongly consistent estimators for the aforementioned statistical quantities with regular sample average estimators being a specific instance of these. We also present an application of the proposed scheme on a widely used model in population biology. Numerical experiments in this framework show that the time-dependent process characteristics as obtained using our algorithm in each case exhibit excellent agreement with exact results. (C) 2010 Elsevier Inc. All rights reserved.

Transport coefficients for the shear dynamo problem at small Reynolds numbers

We build on the formulation developed in S. Sridhar and N. K. Singh J. Fluid Mech. 664, 265 (2010)] and present a theory of the shear dynamo problem for small magnetic and fluid Reynolds numbers, but for arbitrary values of the shear parameter. Specializing to the case of a mean magnetic field that is slowly varying in time, explicit expressions for the transport coefficients alpha(il) and eta(iml) are derived. We prove that when the velocity field is nonhelical, the transport coefficient alpha(il) vanishes. We then consider forced, stochastic dynamics for the incompressible velocity field at low Reynolds number. An exact, explicit solution for the velocity field is derived, and the velocity spectrum tensor is calculated in terms of the Galilean-invariant forcing statistics. We consider forcing statistics that are nonhelical, isotropic, and delta correlated in time, and specialize to the case when the mean field is a function only of the spatial coordinate X-3 and time tau; this reduction is necessary for comparison with the numerical experiments of A. Brandenburg, K. H. Radler, M. Rheinhardt, and P. J. Kapyla Astrophys. J. 676, 740 (2008)]. Explicit expressions are derived for all four components of the magnetic diffusivity tensor eta(ij) (tau). These are used to prove that the shear-current effect cannot be responsible for dynamo action at small Re and Rm, but for all values of the shear parameter.

An iterative auction mechanism for combinatorial exchanges

Combinatorial exchanges are double sided marketplaces with multiple sellers and multiple buyers trading with the help of combinatorial bids. The allocation and other associated problems in such exchanges are known to be among the hardest to solve among all economic mechanisms. In this paper, we develop computationally efficient iterative auction mechanisms for solving combinatorial exchanges. Our mechanisms satisfy Individual-rationality (IR) and budget-nonnegativity (BN) properties. We also show that our method is bounded and convergent. Our numerical experiments show that our algorithm produces good quality solutions and is computationally efficient.

A new parallel overlapped domain decomposition method for nonlinear dynamic finite element analysis

In this paper a new parallel algorithm for nonlinear transient dynamic analysis of large structures has been presented. An unconditionally stable Newmark-beta method (constant average acceleration technique) has been employed for time integration. The proposed parallel algorithm has been devised within the broad framework of domain decomposition techniques. However, unlike most of the existing parallel algorithms (devised for structural dynamic applications) which are basically derived using nonoverlapped domains, the proposed algorithm uses overlapped domains. The parallel overlapped domain decomposition algorithm proposed in this paper has been formulated by splitting the mass, damping and stiffness matrices arises out of finite element discretisation of a given structure. A predictor-corrector scheme has been formulated for iteratively improving the solution in each step. A computer program based on the proposed algorithm has been developed and implemented with message passing interface as software development environment. PARAM-10000 MIMD parallel computer has been used to evaluate the performances. Numerical experiments have been conducted to validate as well as to evaluate the performance of the proposed parallel algorithm. Comparisons have been made with the conventional nonoverlapped domain decomposition algorithms. Numerical studies indicate that the proposed algorithm is superior in performance to the conventional domain decomposition algorithms. (C) 2003 Elsevier Ltd. All rights reserved.

A fast quasi-Newton method for semi-supervised SVM

Due to its wide applicability, semi-supervised learning is an attractive method for using unlabeled data in classification. In this work, we present a semi-supervised support vector classifier that is designed using quasi-Newton method for nonsmooth convex functions. The proposed algorithm is suitable in dealing with very large number of examples and features. Numerical experiments on various benchmark datasets showed that the proposed algorithm is fast and gives improved generalization performance over the existing methods. Further, a non-linear semi-supervised SVM has been proposed based on a multiple label switching scheme. This non-linear semi-supervised SVM is found to converge faster and it is found to improve generalization performance on several benchmark datasets. (C) 2010 Elsevier Ltd. All rights reserved.

Modeling Indian Ocean circulation: Bay of Bengal fresh water plume and the Arabian Sea Mini Warm Pool

The Indian subcontinent divides the north Indian Ocean into two tropical basins, namely the Arabian Sea and the Bay of Bengal. The Arabian Sea has high salinity whereas the salinity of the Bay of Bengal is much lower due to the contrast in freshwater forcing of the two basins. The freshwater received by the Bay in large amounts during the summer monsoon through river discharge is flushed out annually by ocean circulation. After the withdrawal of the summer monsoon, the Ganga – Brahmaputra river plume flows first along the Indian coast and then around Sri Lanka into the Arabian Sea creating a low salinity pool in the southeastern Arabian Sea (SEAS). In the same region, during the pre-monsoon months of February – April, a warm pool, known as the Arabian Sea Mini Warm Pool (ASMWP), which is distinctly warmer than the rest of the Indian Ocean, takes shape. In fact, this is the warmest region in the world oceans during this period. Simulation of the river plume and its movement as well as its implications to thermodynamics has been a challenging problem for models of Indian Ocean. Here we address these issues using an ocean general circulation model – first we show that the model is capable of reproducing fresh plumes in the Bay of Bengal as well as its movement and then we use the model to determine the processes that lead to formation of the ASMWP. Hydrographic observations from the western Bay of Bengal have shown the presence of a fresh plume along the northern part of the Indian coast during summer monsoon. The Indian Ocean model when forced by realistic winds and climatological river discharge reproduces the fresh plume with reasonable accuracy. The fresh plume does not advect along the Indian coast until the end of summer monsoon. The North Bay Monsoon Current, which flows eastward in the northern Bay, separates the low salinity water from the more saline southern parts of the bay and thus plays an important role in the fresh water budget of the Bay of Bengal. The model also reproduces the surge of the fresh-plume along the Indian coast, into the Arabian Sea during northeast monsoon. Mechanisms that lead to the formation of the Arabian Sea Mini Warm Pool are investigated using several numerical experiments. Contrary to the existing theories, we find that salinity effects are not necessary for the formation of the ASMWP. The orographic effects of the Sahyadris (Western Ghats) and resulting reduction in wind speed leads to the formation of the ASMWP. During November – April, the SEAS behave as a low-wind heatdominated regime where the evolution of sea surface temperature is solely determined by atmospheric forcing. In such regions the evolution of surface layer temperature is not dependent on the characteristics of the subsurface ocean such as the barrier layer and temperature inversion.

{C}^0$ penalty methods for the fully nonlinear Monge-Ampère equation

In this paper, we develop and analyze C(0) penalty methods for the fully nonlinear Monge-Ampere equation det(D(2)u) = f in two dimensions. The key idea in designing our methods is to build discretizations such that the resulting discrete linearizations are symmetric, stable, and consistent with the continuous linearization. We are then able to show the well-posedness of the penalty method as well as quasi-optimal error estimates using the Banach fixed-point theorem as our main tool. Numerical experiments are presented which support the theoretical results.

Two-timescale Q-learning Algorithms with an Application to Routing in Networks

We propose two variants of the Q-learning algorithm that (both) use two timescales. One of these updates Q-values of all feasible state-action pairs at each instant while the other updates Q-values of states with actions chosen according to the ‘current ’ randomized policy updates. A sketch of convergence of the algorithms is shown. Finally, numerical experiments using the proposed algorithms for routing on different network topologies are presented and performance comparisons with the regular Q-learning algorithm are shown.

Incentive Compatible Mechanisms for Resource Procurement in Computational Grids with Rational resource Providers

In a computational grid, the presence of grid resource providers who are rational and intelligent could lead to an overall degradation in the efficiency of the grid. In this paper, we design incentive compatible grid resource procurement mechanisms which ensure that the efficiency of the grid is not affected by the rational behavior of resource providers.In particular, we offer three elegant incentive compatible mechanisms for this purpose: (1) G-DSIC (Grid-Dominant Strategy Incentive Compatible) mechanism (2) G-BIC (Grid-Bayesian Nash Incentive Compatible) mechanism (3) G-OPT(Grid-Optimal) mechanism which minimizes the cost to the grid user, satisfying at the same time, (a) Bayesian incentive compatibility and (b) individual rationality. We evaluate the relative merits and demerits of the above three mechanisms using game theoretical analysis and numerical experiments.

Stochastic Algorithms for Discrete Parameter Simulation Optimization

We present two efficient discrete parameter simulation optimization (DPSO) algorithms for the long-run average cost objective. One of these algorithms uses the smoothed functional approximation (SFA) procedure, while the other is based on simultaneous perturbation stochastic approximation (SPSA). The use of SFA for DPSO had not been proposed previously in the literature. Further, both algorithms adopt an interesting technique of random projections that we present here for the first time. We give a proof of convergence of our algorithms. Next, we present detailed numerical experiments on a problem of admission control with dependent service times. We consider two different settings involving parameter sets that have moderate and large sizes, respectively. On the first setting, we also show performance comparisons with the well-studied optimal computing budget allocation (OCBA) algorithm and also the equal allocation algorithm. Note to Practitioners-Even though SPSA and SFA have been devised in the literature for continuous optimization problems, our results indicate that they can be powerful techniques even when they are adapted to discrete optimization settings. OCBA is widely recognized as one of the most powerful methods for discrete optimization when the parameter sets are of small or moderate size. On a setting involving a parameter set of size 100, we observe that when the computing budget is small, both SPSA and OCBA show similar performance and are better in comparison to SFA, however, as the computing budget is increased, SPSA and SFA show better performance than OCBA. Both our algorithms also show good performance when the parameter set has a size of 10(8). SFA is seen to show the best overall performance. Unlike most other DPSO algorithms in the literature, an advantage with our algorithms is that they are easily implementable regardless of the size of the parameter sets and show good performance in both scenarios.

An Actor-Critic Algorithm for Finite Horizon Markov Decision Processes

We develop a simulation based algorithm for finite horizon Markov decision processes with finite state and finite action space. Illustrative numerical experiments with the proposed algorithm are shown for problems in flow control of communication networks and capacity switching in semiconductor fabrication.

An interior penalty method for a sixth-order elliptic equation

We derive and study a C(0) interior penalty method for a sixth-order elliptic equation on polygonal domains. The method uses the cubic Lagrange finite-element space, which is simple to implement and is readily available in commercial software. After introducing some notation and preliminary results, we provide a detailed derivation of the method. We then prove the well-posedness of the method as well as derive quasi-optimal error estimates in the energy norm. The proof is based on replacing Galerkin orthogonality with a posteriori analysis techniques. Using this approach, we are able to obtain a Cea-like lemma with minimal regularity assumptions on the solution. Numerical experiments are presented that support the theoretical findings.

Wavelet based spectral finite element modelling and detection of de-lamination in composite beams

In this paper, a model for composite beam with embedded de-lamination is developed using the wavelet based spectral finite element (WSFE) method particularly for damage detection using wave propagation analysis. The simulated responses are used as surrogate experimental results for the inverse problem of detection of damage using wavelet filtering. The WSFE technique is very similar to the fast fourier transform (FFT) based spectral finite element (FSFE) except that it uses compactly supported Daubechies scaling function approximation in time. Unlike FSFE formulation with periodicity assumption, the wavelet-based method allows imposition of initial values and thus is free from wrap around problems. This helps in analysis of finite length undamped structures, where the FSFE method fails to simulate accurate response. First, numerical experiments are performed to study the effect of de-lamination on the wave propagation characteristics. The responses are simulated for different de-lamination configurations for both broad-band and narrow-band excitations. Next, simulated responses are used for damage detection using wavelet analysis.