Best Isotonic Regressions,

Given a real-valued weighted function on a finite dag, the isotonic regression of , , is unique except when . We are interested in determining a ``best'' isotonic regression for , where by best we mean a regression satisfying stronger properties than merely having minimal norm. One approach is to use strict regression, which is the limit of the best approximation as approaches , and another is lex regression, which is based on lexical ordering of regression errors. For the strict and lex regressions are unique and the same. For , strict is unique, but we show that may not be, and even when it is unique the two limits may not be the same. For , in general neither of the strict and lex regressions are unique, nor do they always have the same set of optimal regressions, but by expanding the objectives of optimization to we show is the same as lex regression. We also give algorithms for computing the best isotonic regression in certain situations.
View on arXiv