TRIPODS Workshop on Deep Learning and Low-Dimensional Structure

The resurgence of deep neural networks has led to revolutionary success across almost all the areas in engineering and science. The connection between deep neural network and low dimensional models emerge at multiple levels: the structural connection between a deep neural network and a sparsifying algorithm has been well observed and acknowledged; one popular application for neural networks involves recognition across low dimensional invariance.

Given these exciting while less exploited connections, the NSF TRIPODS data science centers at Columbia University and Cornell University are organizing a one-day workshop “Low-dimensional Model and Deep Neural Network” on November 22 (Friday) at Davis Auditorium, Columbia University. This workshop aims to bring together experts in machine learning, applied mathematics, signal processing, and optimization, and to stimulate vibrate discussions towards deeper and more explicit understanding of the connections in between.


Schedule

Breakfast: 8:30am - 9:00am

Opening/Welcome Address: 9:00am - 9:10am

Complete Dictionary Learning via L4-Norm Maximization over the Orthogonal Group with Yi Ma 9:10am - 9:55am

Neural networks for signal processing reinvent (and improve) the wheel with Carlos Granda 9:55am - 10:25am


Coffee Break 10:25am - 11:00am


Deep Learning and Operator-Valued Free Probability: Training and Generalization Dynamics in High Dimensions with Jeffery Pennington 11:00am - 11:30am

Geometric Insights into the Convergence of Nonlinear Temporal-Difference Learning with Joan Bruna 11:30am - 12:00pm


Lunch -- 12:00pm - 2:00pm


On the Implicit Bias of Dropout with Rene Vidal 2:00pm - 2:45pm

Beyond Linearization in Neural Networks with Jason Lee 2:45pm - 3:15pm


Coffee Break 3:15pm - 3:40pm


Local Geometry of One-Hidden-Layer Neural Networks with Yuejie Chi 3:40pm - 4:10pm

Predictive Models from Interpolation with Daniel Hsu 4:10pm - 4:40pm