Project description
Deep neural networks in bioimaging applications
Deep neural networks (DNNs) are computational models where many simple processing units work in parallel in interconnected layers. A DNN performs particular tasks through training, learning the strength of connections between units. DNNs show the capability of improving the quality of biomedical image reconstruction. However, the main objection concerns the difficulty to control the Lipschitz constant of current neural architectures, meaning that a small perturbation of the input can result in a huge deviation of the output, negatively impacting image reconstruction. The EU-funded FunLearn project proposes to address this issue with the use of much shallower networks, which are easier to control. The approach relies on functional optimisation to improve learning architectures, and the development of specific neural networks to solve problems in biomedical imaging.
Fields of science
- natural sciencesbiological sciencesbiochemistrybiomolecules
- natural sciencescomputer and information sciencesartificial intelligencemachine learning
- natural sciencesbiological sciencesmolecular biologystructural biology
- natural sciencescomputer and information sciencesartificial intelligencecomputational intelligence
Programme(s)
Topic(s)
Funding Scheme
ERC-ADG - Advanced GrantHost institution
1015 Lausanne
Switzerland
See on map
Beneficiaries (1)
1015 Lausanne
See on map