Showing papers on "Robustness (computer science) published in 1979"
01 Jan 1979
TL;DR: In this article, the authors define robustness as the property of a procedure which renders the answers it gives insensitive to departures, of a kind which occur in practice, from ideal assumptions.
Abstract: Publisher Summary Robustness may be defined as the property of a procedure which renders the answers it gives insensitive to departures, of a kind which occur in practice, from ideal assumptions Since assumptions imply some kind of scientific model, I believe that it is necessary to look at the process of scientific modelling itself to understand the nature of and the need for robust procedures Against such a view it might be urged that some useful robust procedures have been derived empirically without an explicitly stated model However, an empirical procedure implies some unstated model and there is often great virtue in bringing into the open the kind of assumptions that lead to useful methods The need for robust methods seems to be intimately mixed up with the need for simple models This we now discuss
1,304 citations
TL;DR: A heuristic controller for dynamic processes is presented in this paper whose control policy is able to develop and improve automatically and find an application in those complex systems which have been too difficult to control or which in the past have had to rely on the experience of a human operator.
1,106 citations
TL;DR: In this article, a Monte Carlo sampling study is carried out for pair-wise differences as well as a few selected contrasts and the procedures are compared based on the results of this study.
Abstract: Nine procedures for multiple comparisons of means with unequal variances are reviewed. Modifications in some procedures are proposed either for improvement in their performance or easier implementation. A Monte Carlo sampling study is carried out for pair-wise differences as well as a few selected contrasts and the procedures are compared based on the results of this study. Recommendations for the choice of the procedures are given. Robustness of two procedures designed for homogeneous variances under violation of that assumption is also examined in the Monte Carlo study.
205 citations
TL;DR: Chow, Mallet-Paret, and Yorke have recently proposed an algorithm for computing Brouwer fixed points of C^2 maps as discussed by the authors, and a numerical implementation of that algorithm is presented here.
158 citations
TL;DR: In this article, a few robust procedures are mentioned, one of which is motivated by maximum likelihood estimation to make it seem more natural and use of this procedure in regression problems is considered in some detail, and an approximate error structure is stated for the robust estimates of the regression coefficients.
Abstract: Users of statistical packages need to be aware of the influence that outlying data points can have on their statistical analyses. Robust procedures provide formal methods to spot these outliers and reduce their influence. Although a few robust procedures are mentioned in this article, one is emphasized; it is motivated by maximum likelihood estimation to make it seem more natural. Use of this procedure in regression problems is considered in some detail, and an approximate error structure is stated for the robust estimates of the regression coefficients. A few examples are given. A suggestion of how these techniques should be implemented in practice is included.
141 citations
01 Jan 1979
117 citations
01 Jan 1979
TL;DR: In this paper, the authors present a theory and methodology of robust estimation for time series having two distinctive types of outliers. But the authors do not define a robustness metric for the time series.
Abstract: Publisher Summary This chapter presents some theory and methodology of robust estimation for time series having two distinctive types of outliers. Research on robust estimation in the time series context has lagged behind, and perhaps understandably so in view of the increased difficulties imposed by dependency and the considerable diversity in qualitative features of time series data sets. For time series parameter, estimation problems, efficiency robustness, and min–max robustness are concepts directly applicable. Influence curves for parameter estimates may also be defined without special difficulties. A greater care is needed in defining breakdown points as the detailed nature of the failure mechanism may be quite important. A major problem that remains is that of providing an appropriate and workable definition of qualitative robustness in the time series context. For time series, the desire for a complete probabilistic description of either a nearly-Gaussian process with outliers, or the corresponding asymptotic distribution of parameter estimates, will often dictate that one specify more than a single finite-dimensional distribution of the process. It is only in special circumstances that the asymptotic distribution of the estimate will depend only upon a single univariate distribution or a single multivariate distribution.
56 citations
TL;DR: This paper explores the robustness of comparisons for a range of network architectures and demonstrates that network sizing as related to the number of voice channels traversing a particular transmission link, and network topology with respect to theNumber of tandem links utilized in an end-to-end connection can have significant effects on the system performance.
Abstract: Various other studies have already indicated potential advantage of packet-switching over tradiational circuit-switching concepts for voice communications. This paper explores the robustness of such comparisons for a range of network architectures. In particular, it is demonstrated that network sizing as related to the number of voice channels traversing a particular transmission link, and network topology with respect to the number of tandem links utilized in an end-to-end connection can have significant effects on the system performance. Estimates for end-to-end delay and the efficiency of trunk utilization are developed as a function of network topology to provide the insight necessary to identify network design limitations. A number of packet-switching design approaches, ranging from a pure datagram to a circuit-switching-like mode, are also discussed together with their effects on overall performance.
48 citations
TL;DR: In this article, extended optimality criteria for robust designs are applied to response surface problems and methods of calculation are described and the criteria illustrated with several examples; the extended criteria discriminate among designs equivalent by other criteria.
39 citations
TL;DR: In this article, the optimal p-median solutions were computed for six test problems on a network of fortynine demand nodes and compared with solutions from two heuristic algorithms, i.e., the Teitz and Bart heuristic and the Maranzana heuristic.
Abstract: Optimal p-median solutions were computed for six test problems on a network of forty-nine demand nodes and compared with solutions from two heuristic algorithms. Comparison of the optimal solutions with those from the Teitz and Bart heuristic indicates that this heuristic is very robust. Tests of the Maranzana heuristic, however, indicate that it is efficient only for small values of p (numbers of facilities) and that its robustness decreases rapidly as problem size increases.
38 citations
TL;DR: The authors showed that large initial misclassification rates can have serious effects on the error rate of the quadratic discriminant function, and that the effects are more serious for large initialization rates.
Abstract: Initial misclassification can have serious effects on the error rate of the quadratic discriminant function The effects are more serious for large initial misclassification rates, and equal misclassification rates in both groups do not alleviate the problem as in the equal covariance case
TL;DR: Several analyses are described in which robustness considerations have proved relevant and graphic displays as aids to formulating a preliminary model and summary statistics that reduce the influence of outliers are described.
Abstract: We describe several analyses in which robustness considerations have proved relevant. These examples exhibit the importance of (a) graphic displays as aids to formulating a preliminary model; (b) summary statistics that reduce the influence of outliers, that (c) give added opportunities of detecting relationships, and (d) are not unduly sensitive to granularity in the observations; and (e) techniques that pay due attention to anomalies in the data that superficially may appear to be negligible but that can obscure important effects. Finally, we make some general comments on the advantages and disadvantages of robust methodology.
TL;DR: Criteria for evaluating generic cross-impact models: internal consistency, robustness, generality, and clarity are proposed and demonstrated.
TL;DR: In this paper, the robustness properties of a multivariable feedback system under additive and multiplicative perturbations were investigated. But the robust properties of multivariability feedback systems are not directly compared.
Abstract: We prove an inequality that relates the robustness properties of a multivariable feedback system under additive and multiplicative perturbations.
TL;DR: This paper develops and analyzes a quality control algorithm which is suitable for monitoring the performance of direct links via high-volume service observing data and processes a sample of successful as well as unsuccessful call attempts sequentially and in real time to quickly and effectively detect deteriorations in theperformance of a link.
Abstract: The automation of network service observing through message classification software permits a great increase in the rate at which observations can be made, thus allowing the monitoring of network links as well as switching nodes. This paper develops and analyzes a quality control algorithm which is suitable for monitoring the performance of direct (end-office to endoffice) links via high-volume service observing data. This algorithm processes a sample of successful as well as unsuccessful call attempts sequentially and in real time to quickly and effectively detect deteriorations in the performance of a link. Discussion of the performance qualities and the robustness of this algorithm is presented. The discussion is supported by quantitative results.
01 Jan 1979
TL;DR: This chapter presents the study of robustness, an active process of data analysis, and not just a passive evaluation of estimate performance in the Monte Carlo study.
Abstract: Publisher Summary This chapter presents the study of robustness. The Monte Carlo study should be an active process of data analysis, and not just a passive evaluation of estimate performance. In the Monte Carlo, each data set belongs to an inner configuration that is determined by how that data set was generated as part of the raw material for the Monte Carlo. Thus, the raw material for studying estimates in a given finite approximation consists of a list, each element of which combines (1) a data set, (2) information about the inner configuration to which that data set belongs and (3) of instructions of how to use the latter information. In assessing the behavior of an estimate, each element of such a list is used to determine the performance of an estimate, first for the given data set and then for the whole inner configuration that it represents.
TL;DR: In this article, the robustness of prediction error identification methods for approximate models is discussed and is shown to depend on the availability of initial estimates, and a numerically stable realization procedure for sample covariance functions of measured variables, which is based upon a singular value decomposition of an appropriate Hankel matrix.
01 Dec 1979
TL;DR: In this article, the "dead beat" or discrete minimum time control law for sampled data systems is formulated as an output predictive linear equation problem rather than an eigenvalue/eigenvector assignment problem, which allows one to derive very general bounds for closed loop stability under conditions of both model and prediction.
Abstract: The "dead beat" or "discrete minimum time" control law for sampled data systems is formulated as an output predictive linear equation problem rather than an eigenvalue/eigenvector assignment problem. Using the contraction mapping principle and a singular value analysis of this linear equation problem, then allows one to derive very general bounds for closed loop stability under conditions of both model and prediction, large scale errors. Furthermore, a direct design procedure is provided for selection of the "optimal" sample time to maximize closed loop robustness.
TL;DR: The algorithm uses a version of the simplex method of non-linear programming to direct parameter selection for a bank of Kidman filters, thus combining the known convergence characteristics of partitioned adaptive filters with the robustness of this search technique.
Abstract: A new combination of mathematical programming methods and partitioned adaptive filters is proposed for the simultaneous real-time estimation of the parameters and state of an unknown linear system. The algorithm uses a version of the simplex method of non-linear programming to direct parameter selection for a bank of Kidman filters, thus combining the known convergence characteristics of partitioned adaptive filters with the robustness of this search technique. Motivation for the approach is discussed in light of recently published results concerning convergence of the decision function of the partitioned filters, and a simple example is given.
31 Aug 1979
TL;DR: In this article, the authors illustrate how the MTDC control problem can be analyzed from a physical viewpoint, and how the physical insights can be utilized to arrive at a reasonable, robust control design.
Abstract: : The successful application of two terminal dc links suggests that even greater flexibility in power dispatch and enhancement in stability can be realized by a multiterminal dc (MTDC) system embedded in an interconnected ac grid. Several utilities are looking into the possibility of building MTDC systems by adding new terminals to the existing dc links or by connecting dc links that are located in close proximity. It is quite conceivable that MTDC will become a major part of our energy delivery system by the end of this century. The intent of this paper is to illustrate how the MTDC control problem can be analyzed from a physical viewpoint, and how the physical insights can be utilized to arrive at a reasonable, robust control design. The main emphasis is on robustness. Some of the latest results in robustness theory are used in the design stage as well as in analyzing the final design.
TL;DR: In this article, an extension of robustness analysis to incorporate reactive response of other decision makers to the action of the initial decision maker is presented, which has not previously been considered.
Abstract: Robustness analysis has not previously incorporated explicitly the reactive response of other decision makers to the action of the initial decision maker. This note presents an extension of robustness analysis to incorporate this response of the decisional environment.
Posted Content•
TL;DR: In this article, the authors examine the nature of the inconsistency and suggest a general test for misspecification, which can be used to identify the potential inconsistency of OLS models.
Abstract: It is well-known that ordinary least-squares will produce inconsistent estimates of the regression parameters if the dependent variable is censored or truncated. Maximum likelihood estimation with a normality assumption on Tobit and other limited dependent variable models is being employed with increasing frequency to avoid this inconsistency. It is not so commonly acknowledged, however, that such estimates lack robustness : The assumptions required of these models are quite strong and any violation, such as heteroscedasticity or nonnormality, may result in an asymptotic bias as severe as in the naive OLS formulations. But to recognize the potential inconsistency in the face of misspecification without a test for and solution to such misspecification is of little use. The purpose of this paper is to examine the nature of the inconsistency and to suggest a general test for misspecification.
01 Dec 1979
TL;DR: In this article, the authors present a method for estimating bounds on system perturbations in terms of singular values associated with parameter uncertainties and unmodeled dynamics, including some forms of nonlinearities.
Abstract: : This is the final report on a multi-year research program to bridge the gap between modern control theory and practical control system design. The results of the total program are summarized and the last year's results are described in detail. Characterizations of optimal linear controls have been derived, from which guides for selecting the structure of the control system and the weights in the performance index were developed, so that optimal controls meeting conventional design specifications can be designed for single-input systems. For multi-input systems a method of selecting weights for the performance index on the basis of desired closed-loop modal properties and system bandwidth was developed. For such systems it was also found that optimal controls which use state estimators have no guaranteed stability margins, and a method was developed for adjusting the state estimator design so that good margins could be obtained. It was also found that single-loop stability margins in multi-input systems were inadequate measures of robustness, the quality of stability being preserved for a class of perturbations of system characteristics. A new measure of robustness expressed in terms of singular values was developed which is a valid generalization of the concept of stability margins for single-input systems. A method for estimating bounds on system perturbations in terms of singular values associated with parameter uncertainties and unmodeled dynamics, including some forms of nonlinearities, is described in detail in this report. Illustrative examples have been treated in each phase of the program. In this report such an example is presented, demonstrating the se of singular value analysis in design.
[...]
TL;DR: In this article, a self-adapting procedure for the "Model Predictive Heuristic Control" strategy, which was described in a previous communication, was proposed. But this procedure was not applied in our case.
TL;DR: The results presented here indicate that interactive switching between algorithms can increase the robustness and efficiency of the optimization process.
TL;DR: A representation of controllable linear systems is introduced, which permits assigning poles or characteristic parameters to a state feedback system by a matrix multiplication, used as a link between state space and classical parameter plane methods.
Abstract: : A representation of controllable linear systems is introduced, which permits assigning poles or characteristic parameters to a state feedback system by a matrix multiplication This is used as a link between state space and classical parameter plane methods The system representation maps a point in a nxp dimensional parameter space of characteristic parameters into the nxp dimensional parameter space of state feedback gains, where p is the number of actuators For p counts one the coordinates of the characteristic parameter space are the coefficients of the closed loop characteristic polynomial, for p greater than one they are coefficients in a characteristic polynomial matrix and its determinant is the characteristic polynomial By this computationally simple mapping procedure it becomes feasible to map not only a fixed set of eigenvalues but also regions in the s or z plane, in which the eigenvalues shall be located This relaxation of the dynamic specifications permits satisfying other typical design specifications like robustness with respect to sensor and actuator failures, large parameter variations, finite word length implementation, and actuator constraints All tradeoffs between such requirements can be made in the feedback gain space Three examples illustrate the variety of problems which can be tackled with this new tool (Author)
01 Dec 1979
TL;DR: A robust (minimax) quantizer is derived which produces the largest possible guaranteed signal-to-noise ratio over a sizeable class of input distributions (all unimodal distributions with a known moment).
Abstract: This paper considers the design of a quantizer when the probability distribution of the input signal is not known exactly. A robust (minimax) quantizer is derived which produces the largest possible guaranteed signal-to-noise ratio over a sizeable class of input distributions (all unimodal distributions with a known moment). A closed-form solution for the robust quantization levels is possible by using the companding approximation. In many instances, the robust quantizer guarantees a significantly higher signal-to-noise ratio than quantizers designed for the standard distributions (uniform, Gaussian, Laplace.) A similar approach is used to "robustify" the commonly used µ-law and A-law companders.
01 Jan 1979
TL;DR: This paper will be searching for the distribution least favorable for the class of distributions specified by the partial knowledge of the state of nature and correspondingly the statistical procedure providing the greatest lower bound on performance over the admissible distributions.
Abstract: Publisher Summary One of the tasks of an engineer is to adapt ideas and results from other disciplines to his problems. The research of the last decade on statistical robustness provides a course of inquiry we have been attempting to focus on the problems faced by the communications engineer. In this paper we concentrate on two areas where robust ideas have been applied: estimation, and detection. Throughout the paper, we use the minmax notion of robustness. We will be searching for the distribution least favorable for the class of distributions specified by our partial knowledge of the state of nature and correspondingly the statistical procedure providing the greatest lower bound on performance over the admissible distributions.