Undergraduate Seminar: "How computers learn to recognize cats and dogs: an introduction to deep learning and the optimization methods behind the curtain"
Abstract: In this talk, we will experience some simple neural networks to some most advanced state-of-the-art computer vision model (ResNet) in PyTorch (https://github.com/pytorch/pytorch) and observe how they achieve the then-deemed-impossible pattern recognition tasks 20 years ago. Then we will take a sneak peek of what optimization methods involved in training these big models will be featured in Spring 2021 Math 450: Optimization Methods in Machine Learning, as well as how to participate in a Kaggle machine learning competition to boost your resume. To make the experience most enjoyable, it is advised that participants either (1) view the Zoom presentation on a mobile device with resolution density greater than 300 ppi while running the Kaggle cloud kernel shared during the talk on a laptop or (2) watch the presentation and run the Kaggle cloud kernel side by side on a monitor with resolution greater than FHD (1920 x 1080).
Host: Adeli Hutton