Distance to axes and four points (increasing with intensity)
Three different local solvers for computing distances to two points
I am a PhD student at SSRI since march 2010, employed by their subsidiary Kalkulo AS and fully sponsored by industrial means from Statoil and the Research Council of Norway. My research project is created on subjects from the Compound Earth Simulator project of Statoil, which Kalkulo is helping them to develop. I have a MSc in Engineering Mathematics from Lund University, with a specialization in Financial Modelling. Moreover I've been working as a teacher assistant in several courses at the Statistical and Mathematical branch of Lund University, Faculty of Engineering. After graduating in early 2009 I worked as an Analyst/Statistician at Ericsson Consumer Lab, until the end of 2009 when I became a Research Trainee at Kalkulo. My interest in financial mathematics remains strong, and I take any opportunity to discuss the subject over a cup of coffee.
My work is connected to front propagating methods such as Fast Marching-, Fast Sweeping-, and Fast Iterative- methods. All these algorithms solve a specific class of Hamilton Jacobi equations, where the solution can be thought of as a weighted distance from a given start position. The methods are used in a wide range of applications, of which the following are of specific interest:
- Modelling of Geological Folding
- Seismic Forward Modelling
- MR and CT image segmentation
In my latest project I developed an algorithm for 3D fold modeling, called 3D PMM. This algorithm has a high level of parallelism, and was ported to a GPU using Mint with the help of collaborators at UCSD and Mohammed Sourouri. Parts of this work will be presented at EAGE and the ICCS. Mohammed's master thesis on this subject was recently endorsed by gpuscience.com.
I have also created the Semi-Ordered Fast Iterative (SOFI) method for static Hamilton Jacobi equations. This work will was presented at MODSIM 2011.
Previous research projects include the accuracy of numerical stencils used in distance computations. With a minor modification of the local solver the accuracy and efficiency of Eikonal solution methods can be increased. This project is to appear in the Springer journal Computational Geosciences.