Back To Schedule
Monday, October 14 • 3:30pm - 3:45pm
Identification of Kernels in a Convolutional Neural Network: Connections Between Level Set Equation and Deep Learning for Image Segmentation

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!


Two common techniques for image segmentation - level set methods and convolutional neural networks (CNN) - rely on alternating convolutions with nonlinearities to describe image features: neural networks with mean-zero convolution kernels can be treated as upwind finite difference discretizations of differential equations. Such a comparison provides a well-established framework for proving properties of CNNs, such as stability and approximation accuracy. We test this relationship by constructing a level set network, a CNN where forward-propagation is equivalent to solving the level set equation. The level set network achieves comparable segmentation accuracy to solving the level set equation, while not obtaining the accuracy of a CNN. We therefore analyze which convolution filters are present in our CNN, to see whether finite difference stencils are learned during training. We observe certain patterns form in the decoding layers of the network, where kernels cannot be accounted for by finite difference stencils alone.


Jonas Actor

Presenter, Rice University

David Fuentes

UT MD Anderson Cancer Center
avatar for Beatrice Riviere

Beatrice Riviere

Noah Harding Chair and Professor, Rice University

Monday October 14, 2019 3:30pm - 3:45pm CDT
BRC 103

Attendees (1)