George Barbastathis
Massachusetts Institute of Technology
George Barbastathis received the Diploma in Electrical and Computer Engineering in 1993 from the National Technical University of Athens (Εθνικό Μετσόβιο Πολυτεχνείο) and the MSc and PhD degrees in Electrical Engineering in 1994 and 1997, respectively, from the California Institute of Technology (Caltech). After post-doctoral work at the University of Illinois at Urbana-Champaign, he joined the faculty at MIT in 1999, where he is now the Ralph E. and Eloise F. Cross Professor of Manufacturing and Professor of Mechanical Engineering. He has held sabbatical appointments at Harvard University and the University of Michigan – Shanghai Jiao Tong University Joint Institute (密西根交大学院), and has been one of the longest-serving Principal Investigators with the Singapore-MIT Alliance for Research and Technology (SMART). He is member of the Institute of Electrical and Electronics Engineering (IEEE) and the American Mathematical Society (AMS), a Fellow of the Optical Society of America (OSA), which was recently rebranded as Optica, and a Fellow of the Society for Photo Instrumentation Engineering (SPIE). He has served as Associate Editor for the Journal of the Optical Society of America A and the journal Optica, as a committee member, chair and co-chair of numerous conferences, and in several co-founding and consulting capacities for incumbent and start-up companies, as well as law firms.
Presentation Title:
Machine Learning for Computational Imaging: What Did We Learn? What Did the Machines Learn?
Abstract:
For the past decade, my group has been working on machine learning for computational imaging and, more specifically, supervised training of regularizers for ill-posed and ill-conditioned inverse problems. The notion of data-driven regularization has been around since at least the invention of learned sparse codes, or “dictionaries,” by Olshausen and Field in 1996. Learning the code through a deep neural network, as pointed out in 2010 by Gregor and Lecun, narrows the regularizer on-demand. Thus, improvements should be expected in terms of resilience to noise and incomplete measurements. Indeed, circa 2017~18, Unser’s group at EPFL devised a cascaded neural proximal gradient scheme for linear tomography with reduced exposure dose; Ozcan’s group at UCLA showed that a low-NA microscope objective can capture features of comparable fidelity as high-NA; and my group at MIT discovered that a Gerchberg-Saxton-Fienup module followed by a deep neural network can retrieve the phase of the complex field robustly from highly noisy intensity measurements.
Using neural networks for imaging in various contexts is now widespread. What have we learned from these deployments? It is certainly true that neural networks capture priors effectively. It is less clear how optical physics, to the degree that it is involved in machine training and if at all, influences the reconstruction accuracy. I will discuss this topic in the context of more recent work by my group on image-based parameter estimation for dynamical phenomena that are severely undersampled in both space and time as well as severely photon-limited.
Organization Page
- Fields
- SYM-FALL-2026