Obtaining minimum-correlation Latin Hypercube Sampling plans using an IP-based heuristic
نویسنده
چکیده
The objective of Latin Hypercube Sampling is to determine an effective procedure for sampling from a (possibly correlated) multivariate population to estimate the distribution function (or at least a significant number of moments) of a complicated function of its variables. The typical application involves a computer-based model in which it is largely impossible to find a way (closed form or numerical) to do the necessary transformation of variables and where it is expensive to run in terms of computing resources and time. Classical approaches to hypercube sampling have used sophisticated stratified sampling techniques; but such sampiing may provide incorrect measures of the output parameters' variances or covariances due to correlation between the sampling pairs. In this work, we offer a strategy which provides a sampling specification minimizing the sum of the absolute values of the pairwise differences between the true and sampled correlation pairs. We show that optimal plans can be obtained for even small sample sizes. We consider the characteristics of permutation matrices which minimize the sum of correlations between column pairs and then present an effective heuristic for solution. This heuristic generally finds plans which match the correlation structure exactly. When it does not, we provide a hybrid lagrangian/heuristic method, which empirically has found the optimal solution for all cases tested. Z u s a m m e n f a s s u n g . Lateinische Hyperwtirfel Stichprobenverfahren (LHS) dienen dazu, in geeigneter Weise die Verteilungsfunktion (zumindest angenahert) der Funktionswerte einer komplexen (impliziten) Funktion, in Abhangigkeit ihrer Variablenwerte, zu sch~itzen. Anwendungen finden sich in Modellen, in denen erforderliche Variablenumformungen nicht m6glich sind und in denen die Zahl der Simulationsl~ufe aus zeitlichen Grfinden gering zu halten oder fixiert ist. Es stellt sich die Frage, welche Werte in jedem Lauf den Variablen zuzuordnen sind. Herk6mmliche Vorgehensweisen benutzen ausgefeilte, geschichtete Stich* Supported in part by a grant from the National Institute of Standards and Technology (60NANB9D-0974). ** Supported in part by grants from the Office of Naval Research (N00014-90-J-1324) and the Air Force Office of Scientific Research (F49 620-90-C-0022). Correspondence to: C.M. Harris probenverfahren, die jedoch Fehler bei der Bestimmung von Varianz und Kovarianz, aufgrund der Korrelation der Stichprobenpaare, beinhalten k~3nnen. In dieser Arbeit wird eine Methode beschrieben, den absoluten Fehler zwischen dem tats~ichlichen und dem korrelierenden Stichprobenpaar so klein wie m6glich zu halten. Selbst fur kleine Stichprobenumf~inge ktinnen dabei schon optimale Pl~ine erzielt werden. Permutationsmatrizen haben die Eigenschaft, die Summe der Korrelationen zwischen Spaltenpaaren zu minimieren. Die vorgestellte Heuristik ist in der Lage, in allen getesteten Fallen das Optimum zu finden.
منابع مشابه
A conditioned Latin hypercube method for sampling in the presence of ancillary information
This paper presents the conditioned Latin hypercube as a sampling strategy of an area with prior information represented as exhaustive ancillary data. Latin hypercube sampling (LHS) is a stratified random procedure that provides an efficient way of sampling variables from their multivariate distributions. It provides a full coverage of the range of each variable by maximally stratifying the mar...
متن کاملUSING LATIN HYPERCUBE SAMPLING BASED ON THE ANN-HPSOGA MODEL FOR ESTIMATION OF THE CREATION PROBABILITY OF DAMAGED ZONE AROUND UNDERGROUND SPACES
The excavation damaged zone (EDZ) can be defined as a rock zone where the rock properties and conditions have been changed due to the processes related to an excavation. This zone affects the behavior of rock mass surrounding the construction that reduces the stability and safety factor and increase probability of failure of the structure. In this paper, a methodology was examined for computing...
متن کاملCommodity price uncertainty propagation in open-pit mine production planning by Latin hypercube sampling method
Production planning of an open-pit mine is a procedure during which the rock blocks are assigned to different production periods in a way that leads to the highest net present value (NPV) subject to some operational and technical constraints. This process becomes much more complicated by incorporation of the uncertainty existing in the input parameters. The commodity price uncertainty is among ...
متن کاملCentered L2-discrepancy of random sampling and Latin hypercube design, and construction of uniform designs
In this paper properties and construction of designs under a centered version of the L2-discrepancy are analyzed. The theoretic expectation and variance of this discrepancy are derived for random designs and Latin hypercube designs. The expectation and variance of Latin hypercube designs are significantly lower than that of random designs. While in dimension one the unique uniform design is als...
متن کاملAsymptotically Valid Confidence Intervals for Quantiles and Values-at-Risk When Applying Latin Hypercube Sampling
Quantiles, which are also known as values-at-risk in finance, are often used as risk measures. Latin hypercube sampling (LHS) is a variance-reduction technique (VRT) that induces correlation among the generated samples in such a way as to increase efficiency under certain conditions; it can be thought of as an extension of stratified sampling in multiple dimensions. This paper develops asymptot...
متن کامل