Explore advanced computational imaging techniques in this 37-minute lecture by Michael Unser from EPFL, presented at the Institut des Hautes Etudes Scientifiques (IHES). Delve into the remarkable capabilities of deep neural networks in enhancing biomedical image reconstruction quality. Examine the concerns regarding stability and reliability of these tools, and discover iterative reconstruction schemes utilizing stable Lipschitz-constrained networks with learnable activation functions. Investigate not-so-deep neuronal architectures that offer near state-of-the-art performance while providing theoretical guarantees against hallucinations. Learn about a configuration with variational interpretation, featuring components reminiscent of wavelets and multiresolution decompositions introduced by Stéphane Mallat.