Deep Neural Network and Multigrid

4:30PM at LWSN 1142
Prof. Jinchao Xu, Penn State University
Deep Neural Network and Multigrid
Jie Shen

In this talk, I will first give an elementary introduction to some deep learning models and traditional algorithms such as finite element and multigrid methods. I will then explore mathematical relationships between these models and algorithms. Such relationships can be used to study, explain and improve the model structures, mathematical properties and relevant training algorithms for deep neural networks. I will report a new training algorithms that can be used to improve the efficiency of CNNs by significantly reducing the redundancy of the model without losing accuracy. By combining multigrid and deep learning methodologies, I will also present a new convolutional neural network (CNN), known as MgNet, that mathematically unify many existing CNN models and computationally competitive.

Refreshment will be served outside of LWSN 1142 after the talk.