Descripción del proyecto
Redes neuronales profundas en aplicaciones de bioimagenología
Las redes neuronales profundas (RNP) son modelos informáticos donde muchas unidades de procesamiento simples funcionan de forma paralela en capas interconectadas. Una RNP lleva a cabo tareas particulares a través de la formación, en que aprende la resistencia de las conexiones entre las unidades. Las RNP son capaces de mejorar la calidad de la reconstrucción de imágenes biomédicas. Sin embargo, el principal obstáculo tiene que ver con la dificultad para controlar la constante de Lipschitz en las arquitecturas neuronales actuales, lo cual significa que una pequeña perturbación en la información de entrada puede provocar una desviación enorme en los resultados, que acaba afectando negativamente a la reconstrucción de imágenes. En el proyecto FunLearn, financiado con fondos europeos, se propone abordar este problema mediante el uso de redes mucho menos profundas y, por lo tanto, más fáciles de controlar. El método se basa en la optimización funcional para mejorar las arquitecturas de aprendizaje y en el desarrollo de redes neuronales específicas a fin de resolver problemas en la imagenología biomédica.
Objetivo
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.
Ámbito científico
- natural sciencesbiological sciencesbiochemistrybiomolecules
- natural sciencescomputer and information sciencesartificial intelligencemachine learning
- natural sciencesbiological sciencesmolecular biologystructural biology
- natural sciencescomputer and information sciencesartificial intelligencecomputational intelligence
Palabras clave
Programa(s)
Régimen de financiación
ERC-ADG - Advanced GrantInstitución de acogida
1015 Lausanne
Suiza