Professor Ozan Öktem specialises in theory and algorithms for solving severely ill-posed inverse problems with emphasis on tomographic imaging. He worked as an applied mathematician for more than 13 years in industry and returned to academia in 2009. His research combines methods from mathematical analysis, differential geometry, and mathematical statistics with techniques from machine learning. Focus lately has been on combining model based approaches with deep neural networks for uncertainty quantification and task adapted reconstruction in large scale inverse problems. The research is spearheaded by concrete challenges in imaging applications from various scientific fields, like 3D electron and fluorescence microscopy in bioimaging, low-dose clinical CT and spatiotemporal PET/CT, x-ray phase contrast tomography for bioimaging and material sciences, and lately seismic tomography for geophysical prospecting and single particle cryo-EM.

Professor Öktem's Inaugural lecture will outline recent approaches for using (deep) convolutional neural networks to solve a wide range of large-scale inverse problems, such as tomographic image reconstruction. Emphasis is on using neural networks for reconstruction that are domain adapted in the sense that they include physics based models for how data is generated. The talk will also discuss recent developments in using generative adversarial networks for uncertainty quantification in inverse problems.