Lei Zhang, Institute of Natural Sciences, Shanghai Jiao Tong University
Room 306, No.5 Science Building
This work considers mode decomposition as a prototypical pattern recognition problem, which can be addressed from the (a priori distinct) perspectives of numerical approximation, statistical inference and deep learning. Programmable and interpretable regression networks for pattern recognition is introduced to address mode decomposition. It is achieved by assembling elementary modules decomposing and recomposing kernels and data. These elementary steps are repeated across levels of abstraction and interpreted from the equivalent perspectives of optimal recovery, game theory and Gaussian process regression (GPR). The structure of some of these networks share intriguing similarities with convolutional neural networks while being interpretable, programmable and amenable to theoretical analysis. I will use Owhadi’s talk at Oberwolfach to illustrate those points . The paper is in the link.