학위논문 (석사)-- 서울대학교 대학원 : 자연과학대학 통계학과, 2019. 2. 원중호.This paper deals with algorithms for solving the problem of estimating isotonic re- gression functions under partial order constraints. When the cost function is an error square function, this problem becomes a quadratic programming problem, and an effi- cient algorithm using a recursive division method is known. Although generalization has been proposed for the case where the cost function is an arbitrary differentiable con- vex function, there is an error in applying huber loss function, and it cannot applied to cost functions that can not be partially differentiated, such as an absolute loss function and a hinge loss function. In this paper, I show the problems of the generalized-isotonic...