Interests​
  • Data privacy and ownership
  • Statistical-relational machine learning
  • Model interpretability
  • Machine learning for healthcare

Publications

EGGS: A Flexible Approach to Relational Modeling of Social Network Spam. Jonathan Brophy and Daniel Lowd. AAAI 2020 workshop on statistical relational learning in AI (STARAI), 2020. New York, NY.
Collective Classification of Social Network Spam. Jonathan Brophy and Daniel Lowd. AAAI 2017 workshop on artificial intelligence and cyber security (AICS), 2017. San Francisco, CA.

Projects

Project | 01
Medifor (Media Forensics)
Funded by DARPA
2017 - Present​
  • Aim of the project was to better detect manipulated images and videos, decide what type of manipulations they contain, and figure out what part of the image or video has been manipulated.

  • This was a multi-team effort spanning a number of universities and institutions.

  • Our objective was to combine the outputs from a number of forensics algorithms built by other teams to come to a final decision about the manipulation of an image or video.

Project | 02
  • Project aimed at using the rich relational structure of social network data to more effectively detect spammers and spam messages.

  • We developed a generic set of logical rules that can be adjusted to any social network domain and instantiated as probabilistic graphical models, over which joint inference can be performed, classifying all spammers and messages at the same time.

Project | 03
  • An exploratory data analysis of all school shootings from 1774 - 2016.

  • Findings include visualizations for the locations of all shootings, types of weapons used, similarities between shootings and shooters, and statistics about the number of deaths and injuries.

Project | 04
  • A survey of the latest techniques used to process large graphs in a distributed manner.

  • Comparisons between ten different techniques are shown, highlighting the advantages and disadvantages of each in various context settings.

Project | 05
  • Project designed to take in sound from the surroundings of a user and overlay a graphical representation of that sound on a pair of augmented reality glasses.

  • This product was designed for people with hearing issues, allowing them the ability to "see" sound.