About

Haruka Eshima, a second-year PhD student at Okinawa Institute of Science and Technology co-supervised by Prof. Makoto Yamada and Prof. Amedeo Roberto Esposito.

I am interested in statistical inference that remains reliable and privacy-preserving under realistic conditions, such as high dimensionality and heavy tails. My goal is to reveal Privacy-Robustness-Fairness-Utility interactions under such conditions.

• When an ideal source of information is deliberately restricted for privacy or security, what can still be inferred?

• Bounding generalization error in a heavy-tailed setting,

• Information-theoretic limits of training data reconstruction under differentially private training (current project),

• Whether the minority data is neglected in robust training,

etc …

I would eventually like to connect theoretical analyses to identify when and why LLM bias or private data leaks arise in practice.

I am drawn to the Socratic idea of the gadfly—using careful questioning to challenge assumptions that have become too comfortable. I hope my research can play a small part in pointing out failures of the systems we increasingly rely on.

I also like to find theories that explain experimental phenomena for simple neural networks. I analyzed neural network architectures for my Master’s thesis, and connected the theory to experimental results in my first project in PhD.

I enjoy learning Robust Statistics, Information Theory, Convex Analysis, and Optimal Transport. In my free time, I enjoy Karate club activities, Duolingo (French, Chinese, Latin), and colored-pencil drawing.