Groupe d’études et de recherche en analyse des décisions

G-88-33

Minimisation d'une fonction convexe séparable avec contraintes de rapport entre les variables

, et

We present a dual algorithm to minimize a separable convex function under ratio constraints between variables. This minimization problem occurs when the dependent variable of a regression model is discret. The algorithm's complexity is shown to be linear, in number of unidimensional minimizations, for convex functions. The complexity if linear, in number of operations, for the quadratic and linear cases.

, 17 pages