Siiri Rautio defended her PhD thesis Ill-posed microlocal imaging using targeted deep learning at University of Helsinki on April 3, 2025. Her opponent was Professor Elena Loli Piccolomini from University of Bologna.
In her work, Siiri made a remarkable tour in the inverse problems landscape. She created new methods for image deblurring, superresolution, X-ray tomography and Electrical Impedance Tomography (EIT). The central theme in her work is to use modern deep learning for improving reconstructions from unstable data, but applying black-box machine learning only in minimal portions to retain maximal interpretability. Hence “targeted deep learning” in the title.
In the evening party, according to the traditions of the Finnish Inverse Problems Society, I gave Siiri a bespoke doctoral hat. In her case, it is a pink fur hat (100% vegan) comprising at least a zillion flexible fibers. Just as there are tunable parameters in the deep learning neural models she built in her thesis work.
