Deep Learning II

Justin Sirignano, University of Illinois at Urbana-Champaign

Usage Details

Blue Waters Trainee 650, Blue Waters Trainee 651, Blue Waters Trainee 652, Blue Waters Trainee 653, Blue Waters Trainee 654, Blue Waters Trainee 655, Blue Waters Trainee 656, Blue Waters Trainee 657, Blue Waters Trainee 658, Blue Waters Trainee 660, Blue Waters Trainee 662, Blue Waters Trainee 663, Blue Waters Trainee 664, Blue Waters Trainee 659, Blue Waters Trainee 661, Blue Waters Instructor 030, Justin Sirignano

An allocation for the Spring 2018 course “IE 598: Deep Learning II” is requested. This is a graduate-level course (PhD and Masters) at the University of Illinois at Urbana- Champaign (UIUC) which will cover recent exciting advances in deep learning. Deep learning has revolutionized areas such as image, text, and speech recognition. In typical applications, deep neural networks may have millions of parameters and are trained on large datasets. Consequently, it is very computationally expensive to train deep learning models. Fortunately, neural network training can be highly parallelized via GPUs. The allocation on Blue Waters will allow students to implement state-of-the-art deep neural networks for important applications. Without Blue Waters, implementing and learning about these deep neural networks would be impossible.

“IE 598: Deep Learning II” is a project course where students will work on substantial projects for a variety of interesting applications. A prerequisite for the Spring course “IE 598: Deep Learning II” is the Fall course “IE598: Deep Learning”. This prerequisite course comprehensively covers deep learning and students implement deep learning models on Blue Waters. The prerequisite course “IE598: Deep Learning” has been taught in Fall 2016 and 2017 with Blue Waters educational allocations (20,000 and 50,000 hours, respectively). Therefore, students who take “IE 598: Deep Learning II” will already be very familiar with Blue Waters and will be able to immediately begin work on a substantial project. It is expected that some of the projects will develop into conference or journal papers.