Nonparametric Combinatorial Regression for Shape Constrained Modeling

被引:4
作者
Koushanfar, Farinaz [1 ]
Majzoobi, Mehrdad [1 ]
Potkonjak, Miodrag [2 ]
机构
[1] Rice Univ, Dept Elect & Comp Engn, Houston, TX 77005 USA
[2] Univ Calif Los Angeles, Dept Comp Sci, Los Angeles, CA 90095 USA
基金
美国国家科学基金会;
关键词
Convex; isotonic; nonparametric; robust regression; shaped constrained regression; unimodal;
D O I
10.1109/TSP.2009.2028937
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
We introduce a unified approach for calculating nonparametric shape constrained regression. Enforcement of the shape constraint often accounts for the impact of a physical phenomenon or a specific property. It also improves the model's predicability and facilitates subsequent optimizations. The regression models are built by transforming the problem into the combinatorial domain where the shape constraints are imposed by bounding the combinatorial search space. We start by addressing isotonicity shape constraint using a dynamic programming algorithm and demonstrate how the problem can be mapped to the graph combinatorics domain. Next we show how a number of other important shape constraints including unimodality, convexity, limited level set, and limited slope can be addressed using the same framework. The flexibility of proposed framework enables solving the shape constrained regression problem with an arbitrary user-defined error metric. This flexibility is exploited to add robustness against outliers to the model. The algorithms are described in detail and their computational complexity is established. The performance and effectiveness of the shape constrained regression is evaluated on traces of temperature and humidity measurements from a deployed sensor network where a high degree of accuracy and robustness is demonstrated.
引用
收藏
页码:626 / 637
页数:12
相关论文
共 39 条
[1]   A fast scaling algorithm for minimizing separable convex functions subject to chain constraints [J].
Ahuja, RK ;
Orlin, JB .
OPERATIONS RESEARCH, 2001, 49 (05) :784-789
[2]   NEW LOOK AT STATISTICAL-MODEL IDENTIFICATION [J].
AKAIKE, H .
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, 1974, AC19 (06) :716-723
[3]   ROBUST METHOD FOR MULTIPLE LINEAR-REGRESSION [J].
ANDREWS, DF .
TECHNOMETRICS, 1974, 16 (04) :523-531
[4]   Weighted Isotonic Regression under the L1 Norm [J].
Angelov, Stanislav ;
Harb, Boulos ;
Kannan, Sampath ;
Wang, Li-San .
PROCEEDINGS OF THE SEVENTHEENTH ANNUAL ACM-SIAM SYMPOSIUM ON DISCRETE ALGORITHMS, 2006, :783-+
[5]  
[Anonymous], 1988, PROBABILITY MATH STA
[6]   FITTING OF POWER-SERIES, MEANING POLYNOMIALS, ILLUSTRATED ON BAND-SPECTROSCOPIC DATA [J].
BEATON, AE ;
TUKEY, JW .
TECHNOMETRICS, 1974, 16 (02) :147-185
[7]   Linear time isotonic and unimodal regression in the L-1 and L-infinity norms [J].
Boyarshinov, Victor ;
Magdon-Ismail, Malik .
JOURNAL OF DISCRETE ALGORITHMS, 2006, 4 (04) :676-691
[8]   MAXIMUM LIKELIHOOD ESTIMATES OF MONOTONE PARAMETERS [J].
BRUNK, HD .
ANNALS OF MATHEMATICAL STATISTICS, 1955, 26 (04) :607-616
[9]  
Bychkovskiy V, 2003, LECT NOTES COMPUT SC, V2634, P301
[10]  
Corman T. H., INTRO ALGORITHMS