Schedule of talks

  • 2:30pm, Tuesday, September 11th: Dejan Slepčev, DNN: An Introduction for Applied Mathematicians
  • 2:30pm, Tuesday, September 18th: Hayden Schaeffer, An Intro to Convolutional Neural Networks
  • 2:30pm, Tuesday, October 2nd: Linan Zhang, Residual Network
  • 2:30pm, Tuesday, October 9th, Matt Thorpe, GAN
  • 2:30pm, Thursday, October 18th, Matt Thorpe, Wasserstein GAN
  • 2:30pm, Tuesday, November 6th, Linan Zhang, A Short Tutorial on TensorFlow with an Example on AlexNet
    Handout: Tensor Flow Setup Instructions
  • 2:30pm, Tuesday, November 13th, Raghavendra Venkatraman, Connection between deep neural networks and differential equations
  • 2:30pm, Tuesday, December 4th, Yifan Sun, Solving High Dimensional PDEs using Deep Learning
    Abstract: We examine how to approximate the solution of semi-linear parabolic PDE with terminal conditions using deep neural network. We will link the solution to a certain SDE, from which we can sample the trajectories, and the terminal condition used in order to define a loss function. The solution is parametrized by Deep Neural Network (DNN). The loss is minimized over sample paths to optimize the parametrization. The talk is based on the paper by Han, Jentzen and E : Time permitting, we may also take a look at a related paper on using DL to solve optimal stopping problem:


Online resources