Asymptotic optimality of RESTART estimators in highly dependable systems

Abstract We consider a wide class of models that includes the highly reliable Markovian systems (HRMS) often used to represent the evolution of multi-component systems in reliability settings. Repair times and component lifetimes are random variables that follow a general distribution, and the repair service adopts a priority repair rule based on system failure risk. Since crude simulation has proved to be inefficient for highly-dependable systems, the RESTART method is used for the estimation of steady-state unavailability and other reliability measures. In this method, a number of simulation retrials are performed when the process enters regions of the state space where the chance of occurrence of a rare event (e.g., a system failure) is higher. The main difficulty involved in applying this method is finding a suitable function, called the importance function, to define the regions. In this paper we introduce an importance function which, for unbalanced systems, represents a great improvement over the importance function used in previous papers. We also demonstrate the asymptotic optimality of RESTART estimators in these models. Several examples are presented to show the effectiveness of the new approach, and probabilities up to the order of 10-42 are accurately estimated with little computational effort.

An endogenous rate of time preference, the Penrose effect, and dynamic optimality of environmental quality.

In the present paper, the endogenous theory of time preference is extended to analyze those processes of capital accumulation and changes in environmental quality that are dynamically optimum with respect to the intertemporal preference ordering of the representative individual of the society in question. The analysis is carried out within the conceptual framework of the dynamic analysis of environmental quality, as has been developed by a number of economists for specific cases of the fisheries and forestry commons. The duality principles on intertemporal preference ordering and capital accumulation are extended to the situation where processes of capital accumulation are subject to the Penrose effect, which exhibit the marginal decrease in the effect of investment in private and social overhead capital upon the rate at which capital is accumulated. The dynamically optimum time-path of economic activities is characterized by the proportionality of two systems of imputed, or efficient, prices, one associated with the given intertemporal ordering and another associated with processes of accumulation of private and social overhead capital. It is particularly shown that the dynamically optimality of the processes of capital accumulation involving both private and social overhead capital is characterized by the conditions that are identical with those involving private capital, with the role of social overhead capital only indirectly exhibited.

The Optimality of Decision Making During Motor Learning

In our daily lives, we often must predict how well we are going to perform in the future based on an evaluation of our current performance and an assessment of how much we will improve with practice. Such predictions can be used to decide whether to invest our time and energy in learning and, if we opt to invest, what rewards we may gain. This thesis investigated whether people are capable of tracking their own learning (i.e. current and future motor ability) and exploiting that information to make decisions related to task reward. In experiment one, participants performed a target aiming task under a visuomotor rotation such that they initially missed the target but gradually improved. After briefly practicing the task, they were asked to select rewards for hits and misses applied to subsequent performance in the task, where selecting a higher reward for hits came at a cost of receiving a lower reward for misses. We found that participants made decisions that were in the direction of optimal and therefore demonstrated knowledge of future task performance. In experiment two, participants learned a novel target aiming task in which they were rewarded for target hits. Every five trials, they could choose a target size which varied inversely with reward value. Although participants’ decisions deviated from optimal, a model suggested that they took into account both past performance, and predicted future performance, when making their decisions. Together, these experiments suggest that people are capable of tracking their own learning and using that information to make sensible decisions related to reward maximization.

Successive approximation methods for solving optimality problems. USSR.

Translation of 2 articles from the Russian journal.

The n-tuple network: coping with sub-optimality

The subject of this thesis is the n-tuple net.work (RAMnet). The major advantage of RAMnets is their speed and the simplicity with which they can be implemented in parallel hardware. On the other hand, this method is not a universal approximator and the training procedure does not involve the minimisation of a cost function. Hence RAMnets are potentially sub-optimal. It is important to understand the source of this sub-optimality and to develop the analytical tools that allow us to quantify the generalisation cost of using this model for any given data. We view RAMnets as classifiers and function approximators and try to determine how critical their lack of' universality and optimality is. In order to understand better the inherent. restrictions of the model, we review RAMnets showing their relationship to a number of well established general models such as: Associative Memories, Kamerva's Sparse Distributed Memory, Radial Basis Functions, General Regression Networks and Bayesian Classifiers. We then benchmark binary RAMnet. model against 23 other algorithms using real-world data from the StatLog Project. This large scale experimental study indicates that RAMnets are often capable of delivering results which are competitive with those obtained by more sophisticated, computationally expensive rnodels. The Frequency Weighted version is also benchmarked and shown to perform worse than the binary RAMnet for large values of the tuple size n. We demonstrate that the main issues in the Frequency Weighted RAMnets is adequate probability estimation and propose Good-Turing estimates in place of the more commonly used :Maximum Likelihood estimates. Having established the viability of the method numerically, we focus on providillg an analytical framework that allows us to quantify the generalisation cost of RAMnets for a given datasetL. For the classification network we provide a semi-quantitative argument which is based on the notion of Tuple distance. It gives a good indication of whether the network will fail for the given data. A rigorous Bayesian framework with Gaussian process prior assumptions is given for the regression n-tuple net. We show how to calculate the generalisation cost of this net and verify the results numerically for one dimensional noisy interpolation problems. We conclude that the n-tuple method of classification based on memorisation of random features can be a powerful alternative to slower cost driven models. The speed of the method is at the expense of its optimality. RAMnets will fail for certain datasets but the cases when they do so are relatively easy to determine with the analytical tools we provide.

Sufficient Conditions of Optimality for Control Pproblem Governed by Variational Inequalities

* This work was completed while the author was visiting the University of Limoges. Support from the laboratoire “Analyse non-linéaire et Optimisation” is gratefully acknowledged.

Sufficient Second Order Optimality Conditions for C^1 Multiobjective Optimization Problems

2000 Mathematics Subject Classification: Primary 90C29; Secondary 90C30.

Optimality Conditions for D.C. Vector Optimization Problems under D.C. Constraints

2000 Mathematics Subject Classification: Primary 90C29; Secondary 49K30.

Second Order Optimality Conditions in Nonsmooth Unconstrained Optimization

AMS subject classification: 49J52, 90C30.

Viability, Optimality and Stability of Dynamical Systems and Estimation of Convergence of Numerical Schemes

AMS subject classification: 49N35,49N55,65Lxx.

Codifferentiable Mappings with Applications to Vector Optimality

AMS subject classification: Primary 49J52; secondary: 26A27, 90C48, 47N10.

Advances in decision support systems on the endeavour for supply chain optimality

Industry practitioners are seeking to create optimal logistics networks through more efficient decision-making leading to a shift of power from a centralized position to a more decentralized approach. This has led to researchers, exploring with vigor, the application of agent based modeling (ABM) in supply chains and more recently, its impact on decision-making. This paper investigates reasons for the shift to decentralized decision-making and the impact on supply chains. Effective decentralization of decision-making with ABM and hybrid modeling is investigated, observing the methods and potential of achieving optimality.

The Optimality of Decision Making During Motor Learning

In our daily lives, we often must predict how well we are going to perform in the future based on an evaluation of our current performance and an assessment of how much we will improve with practice. Such predictions can be used to decide whether to invest our time and energy in learning and, if we opt to invest, what rewards we may gain. This thesis investigated whether people are capable of tracking their own learning (i.e. current and future motor ability) and exploiting that information to make decisions related to task reward. In experiment one, participants performed a target aiming task under a visuomotor rotation such that they initially missed the target but gradually improved. After briefly practicing the task, they were asked to select rewards for hits and misses applied to subsequent performance in the task, where selecting a higher reward for hits came at a cost of receiving a lower reward for misses. We found that participants made decisions that were in the direction of optimal and therefore demonstrated knowledge of future task performance. In experiment two, participants learned a novel target aiming task in which they were rewarded for target hits. Every five trials, they could choose a target size which varied inversely with reward value. Although participants’ decisions deviated from optimal, a model suggested that they took into account both past performance, and predicted future performance, when making their decisions. Together, these experiments suggest that people are capable of tracking their own learning and using that information to make sensible decisions related to reward maximization.

Modeling corneal surfaces with rational functions for high-speed videokeratoscopy data compression

High-speed videokeratoscopy is an emerging technique that enables study of the corneal surface and tear-film dynamics. Unlike its static predecessor, this new technique results in a very large amount of digital data for which storage needs become significant. We aimed to design a compression technique that would use mathematical functions to parsimoniously fit corneal surface data with a minimum number of coefficients. Since the Zernike polynomial functions that have been traditionally used for modeling corneal surfaces may not necessarily correctly represent given corneal surface data in terms of its optical performance, we introduced the concept of Zernike polynomial-based rational functions. Modeling optimality criteria were employed in terms of both the rms surface error as well as the point spread function cross-correlation. The parameters of approximations were estimated using a nonlinear least-squares procedure based on the Levenberg-Marquardt algorithm. A large number of retrospective videokeratoscopic measurements were used to evaluate the performance of the proposed rational-function-based modeling approach. The results indicate that the rational functions almost always outperform the traditional Zernike polynomial approximations with the same number of coefficients.

Metaheuristic approaches to the hybrid flow shop scheduling problem with a cost-related criterion

In the paper, the flow-shop scheduling problem with parallel machines at each stage (machine center) is studied. For each job its release and due date as well as a processing time for its each operation are given. The scheduling criterion consists of three parts: the total weighted earliness, the total weighted tardiness and the total weighted waiting time. The criterion takes into account the costs of storing semi-manufactured products in the course of production and ready-made products as well as penalties for not meeting the deadlines stated in the conditions of the contract with customer. To solve the problem, three constructive algorithms and three metaheuristics (based one Tabu Search and Simulated Annealing techniques) are developed and experimentally analyzed. All the proposed algorithms operate on the notion of so-called operation processing order, i.e. the order of operations on each machine. We show that the problem of schedule construction on the base of a given operation processing order can be reduced to the linear programming task. We also propose some approximation algorithm for schedule construction and show the conditions of its optimality.