This is a past event
We consider a novel family of pseudo-time marching schemes for a class of nonlinear inverse problems of parameter identification that involve statically acquired and noisy measurement data.
We consider a novel family of pseudo-time marching schemes for a class of nonlinear inverse problems of parameter identification that involve statically acquired and noisy measurement data. The usual route to solving these problems is through an iterative, Tikhonov-regularized Newton scheme that is known to yield solutions sensitively dependent upon the regularization parameter. The pseudo-time marching schemes, presently implemented via direct integration and stochastic filters, replace Newtonian iterations by recursion over pseudo-time steps, which play a nearly self-regularizing role. We verify the effectiveness of the proposed schemes through their applications to a family of problems related to structural health assessment and medical diagnostic imaging of breast tissue models.
- Speaker
- Prof. Debasis Roy (Indian Institute of Science)
- Venue
- FN, G013