Hi All,

 

Here comes the information for our next breakfast seminar.

 

BR
/Liang

 

==============================================================================================

Please join us for the next breakfast seminar sponsored by LTH Profile Area: AI and Digitalization (AID) with Niels Christian Overgaard, Math on Oct. 8 2024 at 09:00-10:00 in E:2311. There will be free breakfast.

 

** Please sign up here by Oct. 2 ***

 

When: Tuesday Oct. 8^th at 09:00-10:00, E building, E:2311 (second floor)

 

Speaker: Niels Christian Overgaard

 

Title: Finding the Lipschitz constant needed for efficient scene recovery from CASSI measurements

 

Abstract: The linear inverse problem associated with the standard model for hyperspectral image recovery from CASSI measurements is considered. This is a total variation-regularized least squares problem which is usually solved using first-order iterative minimization algorithms, e.g. ISTA or FISTA.  These methods all require as input an appropriate step length. This step length is related to the Lipschitz constant for the gradient of the quadratic loss function, and equals the square of the largest singular value of the measurement map. In applications, this number is usually computed directly as the largest eigenvalue of a huge square matrix - a compuation that sometimes becomes a bottleneck in an otherwise optimized algorithm. We have effectively eliminate this bottleneck for CASSI reconstructions by showing how the Lipschitz constant can be calculated from a square matrix whose size is easily three orders of magnitudes smaller than in the direct approach. 

 

Upcoming breakfast seminar

 

The breakfast seminars are scheduled once per month with the main purpose of triggering multi-disciplinary discussions and potential collaborations. These are the upcoming planned slots:

 

• Nov 5, 09:00-10:00 @ Control
• Dec 3, 09:00-10:00 @ Math

 

Speaker and registration details will be sent out closer to each slot.

 

Welcome!

The AID breakfast seminar coordinators.