Two professors receive an ERC Advanced Grant project funding
© 2021 EPFL
Professors Aleksandra Radenovic and Michaël Unser from EPFL’s School of Engineering have been awarded European Research Council (ERC) Advanced Grants.
Professor Radenovic, head of the Laboratory of Nanoscale Biology (LBEN), was awarded the grant for her 2D-Liquid project, while Professor Unser, head of the Biomedical Imaging Laboratory (LIB), received the grant for his FunLearn proposal (Functional Learning: From Theory to Application in Bioimaging).
The ERC Advanced Grants are given each year to established, leading principal investigators to fund long-term funding for "ground-breaking, high-risk" research projects in any field.
Abstract 2D-Liquid project - Professor Aleksandra Radenovic
In this project, we will introduce a myriad of nanoscopy techniques to investigate solid-liquid interactions taking advantage of either engineered defects or defects already hosted in 2D materials. We will address the pertinent question on the mechanism of reactivity of 2D materials with aqueous electrolytes at ambient conditions. At the start of the project, we will explore defects hosted in two classes of 2D materials: hexagonal boron nitride (h-BN) and transition metal dichalcogenides (TMDs). Our plans are to define strategies to extend defect imaging combined with other characterization approaches to a multitude of 2D materials. In parallel, we will explore the role of interfacial liquid taking advantage of novel nanofluidic platforms termed angstrom slits that will allow fine-tuning the balance between 2D and 3D liquid. To control defect density in 2D materials, we will use approaches based on focused ion beam irradiation with Xenon and Helium ions.
We will adapt, and develop different nanoscopy tools (such as single-molecule localization microscopy (SMLM), single-particle tracking (spt), Point Accumulation for Imaging in Nanoscale Topography (PAINT) Minimal Emission Fluxes Microscopy (MINFLUX) and Scanning Ion Conductance Microscopy (SICM)) to probe the role of the defects on the dynamics of interfacial charges. All nanoscopy modalities used in 2D-LIQUID project can operate in –situ under ambient conditions and are compatible with the probing of defect chemistry, charge dynamics in different pH environments, and under different solvents or solvent mixtures. We believe that obtained insights regarding the role of defects in dynamics of the surface charges will shed light on the water and ion transport through nanopores, nanotubes but also ultimately narrow angstrom slits. Our findings will propel the development of nanofluidics, biosensing, energy harvesting, molecular separation, and other nanoscale technologies that exploit liquid 2D-material interfaces.
We will adapt, and develop different nanoscopy tools (such as single-molecule localization microscopy (SMLM), single-particle tracking (spt), Point Accumulation for Imaging in Nanoscale Topography (PAINT) Minimal Emission Fluxes Microscopy (MINFLUX) and Scanning Ion Conductance Microscopy (SICM)) to probe the role of the defects on the dynamics of interfacial charges. All nanoscopy modalities used in 2D-LIQUID project can operate in –situ under ambient conditions and are compatible with the probing of defect chemistry, charge dynamics in different pH environments, and under different solvents or solvent mixtures. We believe that obtained insights regarding the role of defects in dynamics of the surface charges will shed light on the water and ion transport through nanopores, nanotubes but also ultimately narrow angstrom slits. Our findings will propel the development of nanofluidics, biosensing, energy harvesting, molecular separation, and other nanoscale technologies that exploit liquid 2D-material interfaces.
Abstract FunLearn project -Professor Michaël Unser
Functional Learning: From Theory to Application in Bioimaging
This research program is motivated by the remarkable ability of deep neural networks to improve the quality of biomedical image reconstruction. While the results reported so far are extremely encouraging, serious reservations have been voiced pertaining to the stability of these tools and the extent to which we can trust their output. The main concern is that it is very difficult to control the Lipschitz constant of the current neural architectures. This means that a small perturbation of the input can result in a huge deviation of the output, which can have devastating effects in the context of image reconstruction.
We believe that the remedy lies in the use of much shallower networks, which are easier to control. However, a reduction in the number of layers will degrade the performance, unless we augment the sophistication of the primary modules; in particular, the nonlinear ones. By drawing on our career-long experience with splines, we therefore propose to rely on the powerful tools of functional optimization to improve learning architectures. This will allow us to develop two novel approaches to learning: sparse simplicial splines, and hierarchical spline networks—an extension of the popular deep ReLU neural networks.
In parallel, we shall develop specific neural networks to solve two outstanding problems in biomedical imaging:
- A “best-of-both-worlds” approach to biomedical image reconstruction, involving the stable integration of state-of-the-art physics-based solvers with the new tools of machine learning;
- The 3D reconstruction of the entire manifold of configurations of a biomolecule from a large collection of very low-dose cryo-electron tomograms. This goal, which may be viewed as the Graal of structural biology, has remained elusive so far and calls for an entirely new paradigm for single-particle analysis.
This research program is motivated by the remarkable ability of deep neural networks to improve the quality of biomedical image reconstruction. While the results reported so far are extremely encouraging, serious reservations have been voiced pertaining to the stability of these tools and the extent to which we can trust their output. The main concern is that it is very difficult to control the Lipschitz constant of the current neural architectures. This means that a small perturbation of the input can result in a huge deviation of the output, which can have devastating effects in the context of image reconstruction.
We believe that the remedy lies in the use of much shallower networks, which are easier to control. However, a reduction in the number of layers will degrade the performance, unless we augment the sophistication of the primary modules; in particular, the nonlinear ones. By drawing on our career-long experience with splines, we therefore propose to rely on the powerful tools of functional optimization to improve learning architectures. This will allow us to develop two novel approaches to learning: sparse simplicial splines, and hierarchical spline networks—an extension of the popular deep ReLU neural networks.
In parallel, we shall develop specific neural networks to solve two outstanding problems in biomedical imaging:
- A “best-of-both-worlds” approach to biomedical image reconstruction, involving the stable integration of state-of-the-art physics-based solvers with the new tools of machine learning;
- The 3D reconstruction of the entire manifold of configurations of a biomolecule from a large collection of very low-dose cryo-electron tomograms. This goal, which may be viewed as the Graal of structural biology, has remained elusive so far and calls for an entirely new paradigm for single-particle analysis.