Sparsity is popular in statistics and machine learning, because it can avoid overfitting, speed up computations, and facilitate interpretations. In deep learning, however, the full potential of sparsity still needs to be explored. The goal of this project is, therefore, the development of statistical theories for sparse deep learning. We include different types of sparsity, and we account for the intricacies of nonconvex optimization. Besides the theories themselves, the main innovation of this project is the way it interlaces high-dimensional statistics, empirical-process theory, and deep learning. Hence, not only is our research a timely contribution to the mathematical foundations of deep learning, but it is also a seed for further research in statistics, mathematics, and computer science.
DFG Programme
Research Grants
