Biography
Gerard Donahue is a third-year PhD Candidate in Computer Science in the Khoury College of Computer Sciences at Northeastern University, Boston MA. Broadly speaking, his interests lie at the intersection of temporal machine learning, video understanding, deep learning, and algorithms. Gerard is currently a research assistant under Dr. Ehsan Elhamifar where his focus involves discovering mathematical techniques to better train deep models to understand progress and procedural execution of tasks in video data.
Before beginning his doctoral studies at Northeastern, Gerard obtained his B.Sc. in Computer Science from the University of New Hampshire (UNH), where he earned Magna Cum Laude upon graduation. At UNH, Gerard worked in the realm of Inverse Reinforcement Learning (IRL) under the advisement of Dr. Marek Petrik, which resulted in a workshop paper regarding unbiased estimation of feature counts in IRL.
In the summer of 2022, Gerard worked as a machine learning research intern at Ultra Electronics in Austin, TX. At Ultra, Gerard used his knowledge of reinforcement learning to spearhead research in the field of Target Localization in collaboration with Dr. Guni Sharon at Texas A&M. In 2020, Gerard worked as a software engineer at Intel Corporation.
Research Interest
While Gerard’s research interests span widely, his primary interest is to harness the power of machine learning algorithms to increase the value of modern technology. He sees temporal machine learning as a prime way to accomplish this, where the temporal dimension of time series data can be utilized with nuance, creativity, and innovation. Recently, he has been exploring the likes of Hyperbolic Machine Learning, where hyperbolic space is used to complement the inherent hierarchical nature of certain types of data.
Publications
- Learning to Predict Activity Progress by Self-Supervised Video Alignment
- Gerard Donahue, and Ehsan Elhamifar, ``Learning to Predict Activity Progress by Self-Supervised Video Alignment’’, 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, 2024, pp. 18667-18677, doi: 10.1109/CVPR52733.2024.01766.
Workshop Papers
- Unbiased Efficient Feature Counts for Inverse RL
- Gerard Donahue, Brendan Crowe, Marek Petrik, Daniel S. Brown NeurIPS. Workshop on Safe and Robust Control of Uncertain Systems, 2021.
News
- Achieved PhD Candidacy at Northeastern.
- Obtained Masters in Science at Northeastern.
Contact
If you are interested in my research or would like to just chat, please reach out :)
- Email: donahue [dot] g [at] northeastern.edu
- LinkedIn Profile
“I believe things cannot make themselves impossible.” – Stephen Hawking