fmincon-SQP

Must be the fitness function (aim function~cost function) for SQP algorithm (fmincon) in quadratic form for solution? Are there any requirement for constraint (linear~nonlinear)?

Can you define this options of fmincon function more explicitly , please? It is not easy to understand.

Lssteplength, Firstorderopt Directional derivative Max. Constraint

---------------------------------------------------------------------

lssteplenght: Does it related with step lenght parameter? X=x+?•d ?:step lenght parameter d:search direction max constraint: Does it related with on the value of deviation of x parameter?

Answer 1

[Walter Roberson]

According to the documentation, the lssteplength field of the output is only created for active-set algorithm and not for SQP.

The firstorderopt is also shown as an output field and not as an option.

I do not see anything in the SQP discussion that would require a quadratic form, but I might have overlooked something in the math. The examples for fmincon show non-quadratic functions.

I have not located any information about Directional derivative or Max. Constraint as yet. Neither of them are listed as potential input options for any of the solvers.

[gocmen]

thank you for your answer. fmincon uses a SQP method.the basic idea of SQP is the modelling nonlinear problem into a quadratic sub problem. In this method QP subproblem is solved at each iteration by different algorithm.