Deep Neural Networks as Iterated Function Systems and a Generalization Bound

Jonathan Vacher

Deep neural networks (DNNs) achieve remarkable performance on a wide range of tasks, yet their mathematical analysis remains fragmented: stability and generalization are typically…