Computational Methods for Large-Scale Machine Learning in ImagingMS36

Machine learning has become an essential tool for automatically analyzing imaging data and has already outperformed humans in some image classification tasks. Despite recent progress, there remain enormous challenges when processing large data sets such as image sequences, 3D images, and videos. This mini-symposium presents cutting edge imaging applications of machine learning as well as novel computational approaches for solving large-scale learning problems including advances in stochastic optimization, high-performance computing, and the design of deep neural networks.

Randomized Newton and quasi-Newton methods for large linear least squares problems
Matthias Chung (Virginia Tech)
End-to-end learning of CNN features in in discrete optimization models for motion and stereo
Thomas Pock (Graz University of Technology)
Memory-Optimal Deep Neural Networks
Gitta Kutyniok (Technische Universität Berlin)
A Batch-Incremental Video Background Estimation Model Using Weighted Low-Rank Approximation of Matrices
Aritra Dutta (KAUST)
Stacked U-Nets: Multi-scale neural nets for natural image segmentation
Sohil Shah (University of Maryland)
PDE-based Algorithms for Convolution Neural Networks
Eldad Haber (University of British Columbia)
Maximum Principle Based Algorithms for Deep Learning
Qianxiao Li (Institute of High Performance Computing)
Mass effect in glioblastomas
George Biros (Institute for Computational Engineering and Sciences, University of Texas at Austin)
Matthias Chung (Virginia Tech)
Lars Ruthotto (Department of Mathematics and Computer Science, Emory University)
deep learning, image reconstruction, inverse problems, machine learning, nonlinear optimization, statistical inverse estimation methods