About Me

I am a PhD student in Electrical & Computer Engineering at Georgia Tech, advised by Prof. Mark Davenport. My areas of research include active learning and learning from comparisons such as similarity judgments and user preferences.

Current Research

My research focuses on developing methods for making machine learning more data efficient using the paradigms of active learning and learning from comparisons. Active learning reduces the label complexity of a task by selecting the most informative samples to label. Comparisons are a natural source of feedback because it is easier for people to compare samples than to evaluate each one individually.

I am interested in motivating task agnostic active learning frameworks by casting different problems as learning from comparisons. I also design active methods that account for the biases that influence human responses. In my recent work, I proposed a confidence aware response model for pairwise comparisons, which captures how the reliability of a response depends on the similarity of the compared options. Building on this model, I developed an active query synthesis framework that generates informative queries directly instead of searching through a pool of candidates, which reduces computational cost while matching the sample efficiency of pool-based methods.

I am also interested in how the problem setting and the data modality shape the design of active data selection methods. To study these methods, I have worked on:

• Preference learning: estimate a user’s preferences from pairwise comparisons between items.
• Control parameter tuning: find optimal controller settings from pairwise comparisons between trajectories.
• Metric learning: learn a data representation from similarity comparisons between data points.
• Classification: acquire labels by soliciting responses to similarity comparisons between data points.

Publications

N. Nadagouda, N. Ahad, M. Tucker and M. Davenport, “Active query synthesis for learning from preferences”, Under review. Also presented at Women in Machine Learning Symposium, International Conference on Machine Learning (ICML), July 2026.
[website]

N. Nadagouda, A. Xu and M. Davenport, “Active metric learning and classification using similarity queries”, in Uncertainty in Artificial Intelligence (UAI), August 2023. Also presented at Human in the Loop Learning Workshop, Neural Information Processing Systems (NeurIPS), December 2022.
[video] [poster]

A. McRae, A. Xu, J. Jin, N. Nadagouda, N. Ahad, P. Guan, S. Karnik, and M. Davenport, “Delta Distancing: A Lifting Approach to Localizing Items From User Comparisons”, in Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), May 2022.
[poster]

N. Nadagouda and M. Davenport, “Switched Hawkes Processes”, in Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), June, 2021. Preliminary results presented at Workshop on Recent Developments on Mathematical/Statistical approaches in Data Science (MSDAS), University of Texas at Dallas, June 2019.
[video] [poster]

G. Canal, M. Connor, J. Jin, N. Nadagouda, M. O’Shaughnessy, C. Rozell, and M. Davenport, “The Picasso Algorithm for Bayesian Localization Via Paired Comparisons in a Union of Subspaces Model”, in Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP), May, 2020.

Abstracts

N. Ahad, N. Nadagouda, E. Dyer and M. Davenport, “Active learning for time instant classification”, at Data-centric Machine Learning Research Workshop, International Conference on Machine Learning (ICML), July 2023.
[poster]

Y. Teng, A. Mamuye, E. Mo, K. Zhu, R. Walker, N. Nadagouda, and M. Davenport, “Range-Only Simultaneous Localization and Mapping using Paired Comparisons”, at IEEE Annual Conf. on RFID, April 2021.

Article

N. Nadagouda, Journey of a researcher: Finding pleasure in the pathless woods, American Ceramic Society Bulletin, Student Perspectives, June/July 2020.

Social Network