About Me

I am a graduate student at the University of Pennsylvania. I am supervised by Prof. Lingjie Liu. Before that, I received Bachelor of Science in Computer Science and Mathematics from the University of Wisconsin-Madison, where I worked in Machine Learning and Optimization Theory (MLOPT) Research Group and advised by Prof. Ramya Korlakai Vinayak, Prof. Matthew L. Malloy, and Prof. Steven J. Schrodi.

Research interests

Safe and Robust ML: Safety and robustness entail the model’s resilience against adversarial attacks, unanticipated scenarios, and data inconsistencies. My research endeavors include methods such as uncertainty estimation, calibration, and Out-of-Distribution detection to enhance the safety and robustness of ML systems. Relevant publications are listed below:

• Threshold-Based Auto-Labeling: [NeurIPS’23]

• Human-in-the-Loop Out-of-Distribution Detection: [ICML’22], [Preprint’23]

• Confidence Regions for Multinomial Parameters: [ISIT’22]

Quantifying Uncertainty in Neural Rendering: Neural rendering, despite its remarkable performance in view synthesis, often leaves questions about the reliability of its reconstructions. Incorporating uncertainty estimation ensures that the results are quantifiably trustworthy. My focus is on infusing Bayesian statistics into neural radiance fields.

Interpretable AI in Biological Research: Interpretable AI can unlock insights into feature significance and causality and facilitate discoveries in biology. Check out our work on [ProjectX’2021] which delves into the interpretability of neural networks in predicting diabetes in mice. My interest also extends to models that improve interpretability, such as Bayesian Neural Networks and Graphical Neural Networks.

Publications

Good Data from Bad Models: Foundations of Threshold-based Auto-labeling
Harit Vishwakarma, Heguang Lin, Frederic Sala, Ramya Korlakai Vinayak
Neural Information Processing Systems (NeurIPS), 2023 (Spotlight)
[arXiv]

Geometry of the Minimum Volume Confidence Sets
Heguang Lin, Mengze Li, Daniel Pimentel-Alarcón, Matthew L. Malloy
IEEE International Symposium on Information Theory (ISIT), 2022.
[arXiv] [Video]

Adaptive Out-of-Distribution Detection with Human-in-the-Loop
Heguang Lin*, Harit Vishwakarma*, Ramya Korlakai Vinayak
International Conference on Machine Learning (ICML), Workshop on Human-Machine Collaboration and Teaming 2022.
[PDF] [Video]

Machine Learning for Glucose Prediction to Identify Diabetes-related Metabolic Pathways
Gautam Agarwal*, Collin Frink*, Brain Hu*, Ziling Hu*, Eliot Kim*,Heguang Lin*
ProjectX Undergraduate Machine Learning Research Competition, 2021.
[OpenReview]

Preprints

Human-In-The-Loop Out-Of-Distribution Detection With False Positive Rate Control
Harit Vishwakarma*, Heguang Lin*, Ramya Korlakai Vinayak
Under review, 2023
[Comming Soon]

* Denotes equal contribution.

Hobbies

I play basketball in my free time. I am doing a heavy leg train to learn to dunk. I watch the NBA – 10 years Memphis Grizzlies fan! I enjoy cooking. I have no hesitation in trying different recipes for bubble tea.

Awards

The David Dewitt Undergraduate Scholarship, Computer Science Department, UW–Madison, 2022
Undergraduate Scholarship for Summer Study, UW–Madison, 2020