Mesoscale Reconstruction of Images and Networks Using Tensor Decomposition
Explore mesoscale reconstruction techniques for images and networks using tensor decomposition in this 50-minute lecture by Hanbaek Lyu from the University of Wisconsin-Madison. Delve into a unified framework that utilizes low-rank mesoscale structures, examining how global reconstruction error relates to mesoscale reconstruction. Discover the application of online CP-dictionary learning for multi-modal datasets, which incorporates CP tensor decomposition to efficiently represent inter-modal relationships. Learn about convergence guarantees and computational advantages of these algorithms. The lecture covers topics such as motivation, examples of networks, traditional pipelines, data-driven approaches, low-rank structures, random product graphs, network sampling, and dictionary learning for networks.