DEPARTMENT OF MATHEMATICAL SCIENCES
Faculty Candidate Undergraduate Talk
Minsuk Shin (Harvard University)
TITLE: A Short Introduction to Deep Learning
ABSTRACT: This short course covers a brief history of deep neural network (deep learning) models, its strength, its limitation, and practical applications of deep learning. This course is aim at college students who have taken an undergraduate probability course.
In this course, I will talk about an interesting historical background of deep learning and why deep learning idea attracts engineers and researchers in various applications. An intuitive explanation of neural network model will be provided, and I will discuss why the computation of deep learning is tractable even when the size of data sets are tremendous in a big data scheme. I will also compare deep learning to other machine learning and statistical models, and discuss why the deep-layered neural network models outperforms the other methodologies.
While deep learning models have been successful in many predictive applications, it has some limitations in terms of interpretability problems and the issues for small-sized data sets (less than 1,000). I will briefly talk about these limitations and recent studies to overcome them. Finally, I will introduce some interesting deep learning applications that have not been solved by classical methodologies.
Pizza and Soda will be served at the talk!