Module I, Advanced Algorithms: Optimization and Search Methods
You should understand the slides, work with the material to be read,
and do the homework:
The "problem" with the NEOS-server we used in class has been fixed.
Thanks to Jason Sarich from ANL.
You should use the NEOS-server suite to find the minimum of our simple
function for significant sets of fixed parameters (a,b,c,d,e,f,g,...).
You should specify a simple linear constraint optimization problem with
linear cost function and solve it with the NEOS-server suite.
Use AMPL input, use a simple problem with two variables you can solve
graphically, use a more complex one with at least 4 variables.
You should form a group with at most one other student,
select in accordance with Prof. Arno Formella one of
the proposed algorithms below,
and elaborate a not too short and not too long article
(10 to 20 pages) about the algorithm, including at least the aspects:
description of the algorithm,
main field of application,
advantages and disadvantages compared to other algorithms,
available software/implementations,
critical discussion of their APIs,
and references on the algorithm and its applications.
- Nelder Mead algorithm
- Newton Raphson
- Rodríquez García-Palomares algorithm
- Levenberg Marquardt algorithm
- great deluge algorithm
- local unimodel sampling
slides (will be completed during the course)
|
