| has gloss | eng: In numerical analysis, isotonic regression (IR) involves finding a weighted least-squares fit x\in \BbbR}^n to a vector a\in \BbbR}^n with weights vector w\in \BbbR}^n subject to a set of monotonicity constraints giving a simple or partial order over the variables. The monotonicity constraints define a directed acyclic graph G=(N,E) over the nodes N=1,2,\ldots,n} corresponding to the variables x=x_1,x_2,\ldots,x_n}. Thus, the IR problem where a simple order is defined corresponds to the following quadratic program (QP): |