[Ipopt] Differences between IPOPT and MATLAB optimization toolbox

[Peter Carbonetto]

> What's the difference between I use Matlab optimization toolbox "fmincon"
> function and IPOPT with AMPL to solve a constrained dynamic optimization
> problem.

It appears that the MATLAB implementation is based on the trust region 
method described in Byrd, Gilbert and Nocedal (2000). I wouldn't be 
able to say precisely the advantage of trust region methods over the 
filter-based methods used by Andreas Wachter and company. 

One advantage of the MATLAB implementaiton is that it might be able to 
handle large, sparse systems more effectively because it solves them using 
preconditioned conjugate gradient.

Two advantages of IPOPT: it's open source, and it's free.

Peter Carbonetto
Ph.D. Candidate
Dept. of Computer Science
University of British Columbia