The ATOMI workshop aims at gathering the most recent and advanced results on nonlinear optimization, imaging problems and machine learning. The workshop will look at the intersection of these three fields and their interplay: for example, one can think of inverse problems in imaging science, where classical methods (such as variational regularisation) and more modern techniques (such as Deep Neural Networks) come together within the common framework of optimization theory.

The workshop encompasses three macroareas:

• Optimization for Machine Learning: optimization techniques for classical classification problems and for training Deep Neural Networks.

• Model-Based Optimization: traditional variational methods for imaging, looking towards new optimization methods and/or improvements of existing ones.

• Data-Driven Methods for Imaging: modern techniques for imaging science based on the employment of Deep Neural Networks as regularization functional (Plug&Play methods) or for compensating limited data frameworks (such as limited angle tomography).