Mario Geiger

Latest

• Scaling description of generalization with number of parameters in deep learning
• Jamming transition as a paradigm to understand the loss landscape of deep neural networks