Description

In mathematics and computing, the Levenberg-Marquardt algorithm, also known as the damped least-squares method, provides a numerical solution to the problem of minimizing a function, generally nonlinear, over a space of parameters of the function. These minimization problems arise especially in least squares curve fitting and nonlinear programming.
The LMA interpolates between the Gauss-Newton algorithm and the method of gradient descent.

• You will have to implement a test algorithm in Matlab or Octave,
• adapt the algorithm to an existing parameter estimation tool,
• generate nice looking plots of the time evolution,
• concise description of the theory, your implementation and your numerical results in form of a PDF,
• give a presentation in our group talk.

Size and difficulty of the project

This project is suited for one or two students (80 hours per person) as an advanced software practical. The software practical can be carried out in the PC pool of IWR or on your own PC / notebook. We will meet regularly to discuss your progress and define your next points of work. The practical shall be completed by the end of this semester.

Requirements

• Good knowledge of the software package Matlab,
• knowledge of the programming language Fortran is helpful,
• linear algebra

Contact

Dipl.-Math. Robert Kircheis
Interdisciplinary Center for Scientific Computing (IWR)
Im Neuenheimer Feld 368
Raum 402
Universität Heidelberg

e-mail: robert.kircheis@iwr.uni-heidelberg.de
Room : INF 368 (IWR), R 402