Aarshvi Gajjar
I am a fourth-year CS Ph.D. student at NYU, where I am affiliated with the Theoretical Computer Science Group. My advisors are Christopher Musco and Chinmay Hegde.
Previously, I completed my masters from UMass Amherst, where I did my thesis with Cameron Musco. I also worked as a Strat at Goldman Sachs and earned my B.Tech. from IIIT Hyderabad.
Research
An important challenge in machine learning is determining how to acquire information effectively by strategically collecting data to improve performance.
My research focuses on developing theoretically sound and (ideally) practical algorithms for this setting. This includes areas like active learning, efficient exploration, optimal design, decision making.
Problems I am currently thinking about:
- Active regression under known or uncertain constraints.
- Parallels between active classification and active regression -- can techniques from one inform adaptive strategies in another?
- Sample and computationally efficient finetuning methods.
Publications
Agnostic Active Learning of Single Index Models with Linear Sample Complexity†Aarshvi Gajjar, †Wai Ming Tai, †Xingyu Xu, Chinmay Hegde, Christopher Musco and Yi Li
COLT, 2024
Associated poster for minisymposium on Scientific ML for Scarce Data, SIAM MDS24
Preliminary version: Adaptive Experimental Design and Active Learning Workshop, NeurIPS, 2023
Active Learning for Single Neuron Models with Lipschitz
Non-Linearities
Aarshvi Gajjar, Chinmay Hegde and Christopher Musco
AISTATS, 2023
Preliminary version: selected as Spotlight at DLDE Workshop , NeurIPS 2022
Subspace Embeddings under Nonlinear Transformations
Aarshvi Gajjar, Cameron Musco
ALT, 2021
Authors are listed alphabetically, except for those marked with †, indicating equal contribution.
Contact: [first-name]@nyu.edu.