Model-Based Deep Learning: Deep Equilibrium Models and Proximal Gradient Maps

Explore advanced concepts in deep learning and signal processing through this webinar featuring Prof. Mathews Jacob from the University of Iowa. Delve into model-based deep learning, deep equilibrium models, and their applications in various fields. Examine standard propagation techniques and their limitations, then discover alternative approaches such as proximal gradient maps and monotone operators. Investigate the importance of adversarial robustness, unique fixed points, and safety in deep learning models. Learn about undersample data challenges and solutions, including deep image priors and deep generator models. Gain insights into practical applications in multislice imaging, speech processing, and MRI, with a focus on motion-compensated image recovery techniques.

Introduction
ModelBased Deep Learning
Deep Equilibrium Models
Standard Propagation
Propagation Limitations
Proximal Gradient Maps
PSY Points
Monotone Operators
Alternate proximal gradient algorithm
monotone operator
adversarial robustness
unique fixed points
safety and robustness
enforcement
undersample data
undersample measurements
undersampling patterns
deep image prior
large scale
nonlinear manifold
neural network
deep generator model
multislice imaging
speechMRI
Motion compensated image recovery