Meet DTU Compute: Per Christian Hansen

At DTU Compute we have the lecture series called “Meet DTU Compute” with the intention of giving everybody an opportunity to meet DTU Compute and get introductions to hot research topics.

The next Meet DTU Compute lecture will be given by Per Christian Hansen on Wednesday, December 18 in room 303a/041 at 12:30 - 14.00. Per Christian will talk about “Inverse Problems – Do the Impossible, Solve the Unsolvable “. The abstract can be found below.

Program:

12:30-13:00 Lunch + networking
13:00-14:00 Meet DTU Compute lecture, see details below

 
Registration if necessary due to catering.
Please register by 16 December via this link.

Speaker: Professor Per Christian Hansen,  Scientific Computing, DTU Compute

Title: Inverse Problems – Do the Impossible, Solve the Unsolvable

Abstract: Inverse problems” is a term used in mathematics to describe a range of techniques where collections of measurements are used to compute the phenomenon or object that gave rise to the data. In this way, we reconstruct hidden information from accessible data. Image deblurring is a well-known inverse problem where we compute the sharp image on the bases of a blurred image and a mathematical model of the blurring (with applications in, e.g., astronomy and biometrics).  Tomographic reconstruction is another inverse problem: when a signal (e.g., electricity, soundwaves, X-rays or neutrons) passes through an object, then measurements of their attenuations give information about the object’s interior (as used in, e.g., X-ray medical imaging and industrial inspection). In spite of the fact that we encounter inverse problems every day, their solution poses a number of mathematical challenges and problems.  For one thing, computer solution of inverse problems requires the solution of millions of equations – which must be done fast and in a stable way. Moreover, the presence of noise and data errors means that, in a strict mathematical sense, the inverse problem does not have a solution. In a nutshell, the challenge is how to solve large computational problems associated with problems that do not have a solution? In this talk, Per Christian Hansen will demonstrate how we use mathematics to deal with these issues in a rigorous way, such that we arrive at well-defined and stable solutions that can be computed efficiently – in spite of the errors and noise in the data. Among the techniques we use are regularization methods, sparsity priors and regularizing iterations.

Everybody is welcome to attend!

Time

Wed 18 Dec 19
12:30 - 14:00

Organizer

DTU Compute

Where

DTU Lyngby Campus, Building 303Aa/Aud.041