Thursday, August 14, 2025 - 3:30 p.m. to 4:30 p.m.
Chase Ocean Engineering Lab -and- Teams
Full title: What should the machine learn? Incorporating geometric knowledge into underwater acoustic feature representations.
Speaker: Dr. Ananya Sen Gupta, University of Iowa
Abstract: Sonar target recognition and shallow water communications are critically important for naval situational awareness and environmental sustainability. Testing popular machine learning techniques over underwater acoustic data is especially interesting as they provide the exciting opportunity to seek new scientific knowledge in a realm dominated by complex, dynamic, unpredictable, and often unknown factors. The underwater acoustic realm also challenges emerging feature engineering and knowledge discovery techniques due to the dominance of non-linear time-varying overlap across important multi-dimensional features. Such overlap leads to feature dictionaries that are difficult to track or quantify, e.g., the persistent target features may occupy a latent but changing subspace or occluded against features from a structured background.
In this talk, I will discuss these challenges commonly encountered in underwater acoustics using case studies from field experiments as well as physics-driven simulations in this domain to provide robust ground truths. I will provide practical examples from my research across several disciplines, discuss how to implement key geometric ideas into well-known AI frameworks, and explain how we can use the powerful combination of topology and geometric braid theory to quantify what makes a machine-discovered feature informative. If time permits, I will mention some other applications of my feature engineering work such as oil spill pollution studies and space plasma informatics.
Teams link and info:
Teams link
Meeting ID: 299 598 868 122 4
Passcode: vQ24tV9B
________________________________________
Dial in by phone
+1 332-249-0712,,993825539# United States, New York City
Find a local number
Phone conference ID: 993 825 539#
All are welcome!
