MDI Workshops Explore Non-Traditional Data Analysis Throughout Spring Semester
Written by Carrie McDonald, MDI Journalism Intern
This spring, the Massive Data Institute’s (MDI) workshops followed the central theme of non-traditional data analysis. Each month, an MDI Fellow led participants through different methods of analysis from network to temporal, spatial, and image.
Led by our Postdoctoral and Data Science Fellows, MDI hosts monthly workshops every semester to introduce all Georgetown community members — faculty, students, and staff — to cutting-edge data analysis methods, programming paradigms, and emerging technologies. This year’s theme of non-traditional data analysis aimed to expose participants to new methods that can be applied to real-world data that they likely have not encountered in a standard classroom setting.
“We looked for a theme that connected the MDI Fellows research areas while bringing new topics to the Georgetown community,” MDI Director Lisa Singh said. “The ability to use innovative types of data for research is no longer a nice-to-have but a necessity.”
MDI Fellow Dr. Helge Marahrens added that non-traditional methods of analysis are key to all the fellows’ work.
The fact that all four of us work with such a diverse range of data highlights that nontraditional is not the exception, it’s the norm
Helge Marahrens, MDI Fellow
Marahrens kicked off the semester in January with his workshop exploring network methods for an interconnected world. Marahrens introduced attendees to the basic tools of network analysis and more advanced machine learning methods for link prediction, all with the aim of analyzing important interdependencies, such as in social networks.
For the second workshop, MDI Fellow Dr. Nathan Wycoff guided participants through methods for temporal analysis in February. During this workshop, titled “Timely Data Analysis: Modeling and Exploiting Temporal Correlation,” Wycoff taught statistical techniques commonly utilized by forecasters, covering fundamental time series models like ARIMA models and the newest machine learning tools leveraging neural networks. These methods are essential for public policy because key decision-makers often use present-day trends to make the best prediction for optimal business or policy outcomes.
The penultimate workshop of the semester concentrated on spatial data analysis. MDI Fellow Dr. Le Bao led attendees through multiple spatial data analysis methods vital to understanding social, political, economic, and natural occurrences, such as point pattern analysis, spatial interpolation, and regression and inference with spatial data.
MDI Fellow Dr. Rupayan Mallick closed out the semester with the final workshop, “Next Generation Generative Models for Vision-Related Tasks.” Building off the prior workshops, Mallick added his perspective on high-quality image generation and synthesis. Aiming to leave participants with a high-level understanding of generative models, Mallick introduced both generative and diffusion models, ultimately using them to highlight examples from applications such as inpainting and adding photorealistic effects.
Mallick also discussed the ethical implications of these methods, saying that “these are generally heavy models to train, so although they are interesting to see and use, there are ethical and environmental concerns.” For example, Mallick noted the difficulty researchers sometimes face with knowing from where the training data for large models originates, and whether it was ethically sourced. Furthermore, Mallick explained environmental concerns inherent to using high computational resources such as a graphics processing unit (GPU).
Overall, the fellows hope that the Georgetown community learned from each of the workshops and will be able to apply their knowledge in their future endeavors.
“We’ve enjoyed the many lively discussions and hope that attendees can use these methods and insights in their work, be it in academia, government, or industry,” Marahrens said.
MDI’s workshop series will return in the next academic year.
