Over 2000 years since Greek mathematician Euclid revolutionized geometry, Justin Solomon of MIT’s Electrical Engineering and Computer Science Department and the Computer Science and Artificial Intelligence Laboratory, is using modern geometric techniques to solve complex problems in various fields. From machine learning to autonomous vehicle recognition systems, Solomon integrates geometry with statistical research for groundbreaking innovations.
Solomon uses geometric techniques to discern the structure and relations between datasets in machine learning, highlighting possible impacts on model performances. This widens the understanding of geometry applications, pushing beyond traditional 2-dimensional and 3-dimensional geometric data analysis. The utilization extends to generating better AI models by mapping datasets to glean more profound insights into their structures.
Having had an early interest in computer graphics, Solomon leveraged opportunities to intern in research labs and Pixar Animation Studios, developing algorithms for 3D face recognition and physical simulations for animated movies. His association with Pixar continued through his graduate school years, fostering his exploration of the unique mathematical challenges in the graphics field.
During his PhD at Stanford, he ventured into optimal transport research, a principle seeking efficient distribution from one point to another at minimal cost. Originally aimed at computer graphics applications, this research unexpectedly expanded into other fields and now makes up the structure of his research group at MIT.
Drawn to MIT for its opportunity to engage in practical research with brilliant students and colleagues, Solomon has prioritized making geometric research accessible to diverse students who may otherwise lack the exposure. He established the Summer Geometry Initiative, a paid research program engaging undergraduates – primarily from underrepresented backgrounds – in hands-on geometric research.
The initiative has yielded impressive results, altering the structure of PhD student demographics not only at MIT but other institutions as well. Solomon believes the growing list of issues that can be tackled with geometric techniques in machine learning and statistics necessitates a diverse field of researchers who bring fresh perspectives.
Currently, Solomon aims to apply geometric tools for improving unsupervised machine learning models. These models, he explains, need to recognize patterns without labeled training data. He suggests that sophisticated models incorporating geometric insight and inference from data can assist computers in understanding complex, unlabeled 3D scenes more efficiently.
Away from his research work, Solomon, an avid musician, delights in playing classical music on the piano or cello. He sees a connection between his musical interests and his research, both requiring analytical thought and integrating well with his work in computer graphics.
In conclusion, Justin Solomon is evolving geometric techniques for modern applications, contributing to the field of machine learning, computer graphics, and autonomous vehicle systems. Simultaneously, he remains committed to broadening access to geometric research through his summer initiative, championing diversity in the field.
