优化问题的一个简要概述(MIT讲义).pdfVIP

A Brief Overview of Optimization Problems� Steven G. Johnson� MIT course 18.335, Fall 2008� Why optimization?� •� In some sense, all engineering design is optimization: choosing design parameters to improve some objective� •� Much of data analysis is also optimization: extracting some model parameters from data while minimizing some error measure (e.g. fitting)� •� Most business decisions = optimization: varying some decision parameters to maximize profit (e.g. investment portfolios, supply chains, etc.)� A general optimization problem� minimize an objective functionf � min f 0 (x ) 0 x �n with respect to n design parameters x� (also called decision parameters, optimization variables, etc.)� — note that maximizing g(x)� corresponds to f 0 (x) = –g(x)� subject to m constraints� f (x ) � 0 note that an equality constraint� i h(x) = 0� i = 1,2,…,m yields two inequality constraints� f i(x) = h(x) andf i+1(x) = –h(x)� (although, in practical algorithms, equality constraints �typically require special handling)� x is a feasible point if it� satisfies all the constraints� feasible r egion = set of all feasible x� Important considerations� •� Global versus local optimization� •� Convex vs. non-convex optimization� •� Unconstrained or box-cons