Lagrange multipliers with inequality constraints

72 Views Asked by At

When solving non-linear optimization with inequality constraints, one method seems to be to divide the problem in two parts and solve it inside the boundary and on the boundary. My question is: why do we need to make the distinction between the boundary and the inside? Also, why do we have to optimize on the inside by setting the gradient equal to zero, and on the boundary by using Lagrange multipliers?