Machine learning is one of those hot topics in Computer Science. At Notre Dame, we are at the forefront of leveraging new learning techniques for applications in facial biometrics and media forensics. With the renaissance of deep learning, the landscape of these fields has completely changed from 3 years ago when I started my work as a PhD student. Why do people think deep learning is so great? What makes it better than previous methods? What even is deep learning? What applications can it be good for? Are deep neural networks all they are chalked up to be? In this talk, I will attempt to answer these questions from the perspective of a PhD student researcher who is tumbling along with the current trends.
Being a PhD student in a field like computer science can be tough sometimes, due to the high turnover rate of research trends. But don't let that discourage you, because the research can also extremely rewarding. Additionally, there is a lot more to being a graduate student than just the research. Along with covering these topics on deep learning, this talk will also be a conversation on what it is like to live and work as a graduate student in this field, including (but not limited to) logistics such as admission applications, funding, laboratories, advisors, PI's, and even housing.
When moving on to graduate studies after college, there can be a lot of confusing and frustrating questions about the process that aren't always well answered. If you are considering post-graduate studies, I hope this talk will serve to alleviate some of these questions and concerns!
Joel Brogan graduated from Hope College in 2014 with a BS in Electrical Engineering and a computer science minor. He began his work as a PhD student at the University of Notre Dame in fall of 2014. There, he works in the Computer Vision Research Laboratory, studying machine learning and computer vision techniques for biometric modalities, such as fingerprint, iris and face recognition.
Joel's research pertains specifically to training methods for deep learning architectures, such as convolutional neural networks, for use in applied facial biometric systems. He studies how to train these networks in faster, more accurate, and more generalizable ways. His work aims to see better performance and less error when these methods are deployed in active commercial or government environments.