ABSTRACT
The method of ordinary least squares approximation is not resistant to data points that cause a disproportionate influence in the fit. When outliers are known to exist in the data, robust estimation algorithms are preferred. However, the performance of most robust estimation algorithms degrades in higher dimensions due to factorial complexity and sparse data. A new polynomial time algorithm RIPPLE has been developed to produce robust estimations of data obtained from piecewise linear functions. This paper presents a comparison between the new algorithm RIPPLE and a standard robust estimation algorithm LMS.
- M. W. Berry and K. S. Minser, "Algorithm 798: High-dimensional interpolation using the modified Shepard method," ACM Transactions on Mathematical Software, vol. 25, no. 3, pp. 353--366, 1999. Google ScholarDigital Library
- P. J. Besl, J. B. Birch, and L. T. Watson, "Robust window operators," Machine Vision and Applications, vol. 2, no. 4, pp. 179--191, 1989.Google ScholarCross Ref
- M. A. Fischler and R. C. Bolles, "Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography," Comm. of the ACM, vol. 24, pp. 381--395, 1981. Google ScholarDigital Library
- R. Franke and G. Neilson, "Smooth interpolation of large sets of scattered data," J. Numer. Methods Eng., vol. 15, pp. 1691--1704, 1980.Google ScholarCross Ref
- V. Gantovnik, Z. Gürdal, and L. T. Watson, "Linear Shepard interpolation for high dimensional piecewise smooth functions," in Proc. 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conf., CD-ROM, Albany, NY, 2004, 15 pages.Google Scholar
- M. A. Iyer and L. T. Watson, "An interpolation method for high dimensional scattered data," in Proc. 2006 Spring Simulation Multiconf., Business and Industry Symp., J. A. Hamilton, Jr., R. MacDonald, and M. J. Chinni (eds.), Soc. for Modeling and Simulation Internat., San Diego, CA, 2006, pp. 217--222.Google Scholar
- R. Myers, "Classical and Modern Regression with Applications," Duxbury Press, 2nd edition, 2000.Google Scholar
- R. J. Renka, "Multivariate interpolation of large sets of scattered data," ACM Transactions on Mathematical Software, vol. 14, no. 2, pp. 139--148, 1988. Google ScholarDigital Library
- P. J. Rousseeuw, "Least median of squares regression," Journal of American Statistical Association, vol. 79, no. 388, pp. 871--880, 1984.Google ScholarCross Ref
- D. Shepard, "A two-dimensional interpolation function for irregularly spaced data," Proceedings of the 1968 23rd ACM national conference, pp. 517--523, 1968. Google ScholarDigital Library
Index Terms
- A performance comparison of RIPPLE and LMS for high dimensional piecewise linear functions
Recommendations
A performance comparison of piecewise linear estimation methods
SpringSim '08: Proceedings of the 2008 Spring simulation multiconferenceThe response functions in many engineering problems are piecewise smooth functions, although globally they can be highly nonlinear. The linear Shepard algorithm (a moving window weighted least squares method based on linear functions) usually creates ...
Pattern Recognition by Piecewise Linear Discriminant Functions
A piecewise linear function is represented in terms of a set of linear functions through the use of the maximum and minimum functions. A procedure for finding piecewise linear discriminant functions for pattern recognition is described. The procedure ...
High-dimensional robust regression with L q -loss functions
AbstractRobust procedures in high-dimensional regression are important because outliers are often present in data. For data with heavy-tailed errors, quantile regression and least absolute deviation regression methods have been widely used ...
Comments