Constrained Optimization

An optimization problem becomes more complicated if it is constrained, i.e., the arguments #tex2html_wrap_inline8272# of the objective function #tex2html_wrap_inline8274# are subject to (abbreviated as s.t. in the following) certain equality and/or inequality constraints in terms of what values they can take. Such a generic constrained optimization problem can be formulated as:

#math748#   #tex2html_wrap_indisplay19729# (151)
The set composed of all values of #tex2html_wrap_inline8276# that satisfy the #tex2html_wrap_inline8278# constraints is called the <#3625#>feasible region<#3625#> of the problem. The goal is to find a point #tex2html_wrap_inline8280# in the feasible region at which #tex2html_wrap_inline8282# is an extremum.

In the following we will first consider the most general case where the objective function #tex2html_wrap_inline8284# is nonlinear and so are the constraint functions #tex2html_wrap_inline8286# and #tex2html_wrap_inline8288#. The processing of solving such a problem is called the <#3631#>nonlinear programming (NLP)<#3631#>. Here <#3632#>programming<#3632#> means the mathematical precedure for solving the optimization problems. We will consider the general method of Lagrange multiplier and a set of conditions all solutions of such problems must satisfy.

We will then consider some specific algorithms for two special cases: <#3633#>linear programming (LP)<#3633#> in Section #LinearProgramming#3634> for problems where the objective function and all constraint functions are linear (the feasible region is a polytope), and then <#3635#>quadratic programming (QP)<#3635#> in Section #quadraticProgram#3636> for problems where the objective function is quadratic while the constraints are linear.