Chiraag Kaushik

About Me

I'm a fourth-year PhD student at the School of Electrical and Computer Engineering and the Center for Machine Learning at Georgia Tech. I am fortunate to be advised by Professors Vidya Muthukumar and Justin Romberg. I completed my BS in Electrical and Computer Engineering in May 2021 from Rice University. My work is generously supported by the NSF Graduate Research Fellowship.

My research interests span high-dimensional and overparameterized machine learning and signal processing. In particular, I am interested in furthering scientific and mathematical understanding of commonly observed empirical phenomena in modern machine learning systems. This has led me to study

• Generalization and robustness, especially in the context of typically ad-hoc training procedures like data augmentation and in modern representation learning
• Feature learning, or the ability of modern ML architectures to leverage low-dimensional structure in data
• Models that interpolate, or perfectly fit, training data

In my free time, I enjoy reading, running, and listening to/practicing Indian classical music.

Publications

Preprints

• MANGO: Disentangled Image Transformation Manifolds with Grouped Operators
  Brighton Ancelin, Yenho Chen, Peimeng Guan, Chiraag Kaushik, Belen Martin-Urcelay, Alex Saad-Falcon, Nakul Singh
  [arxiv]
  
  

Conference papers

• Precise asymptotics of reweighted least-squares algorithms for linear diagonal networks
  Chiraag Kaushik, Justin Romberg, and Vidya Muthukumar
  NeurIPS 2024
  [arxiv]
  
  
• Balanced Data, Imbalanced Spectra: Unveiling Class Disparities with Spectral Imbalance
  Chiraag Kaushik (equal cont.), Ran Liu (equal cont.), Chi-Heng Lin, Amrit Khera, Matthew Jin, Wenrui Ma, Vidya Muthukumar, and Eva L. Dyer
  ICML 2024
  [arxiv]
  
  
• Network topology change-point detection from graph signals with prior spectral signatures
  Chiraag Kaushik, T.M. Roddenberry, and Santiago Segarra
  ICASSP 2021
  [arxiv][code]
  
  

Journal papers

• New equivalences between interpolation and SVMs: Kernels and structured features
  Chiraag Kaushik, Andrew McRae, Mark Davenport, and Vidya Muthukumar
  SIMODS 2024
  [arxiv]
  
  
• The good, the bad and the ugly sides of data augmentation: An implicit spectral regularization perspective
  Chi-Heng Lin, Chiraag Kaushik, Eva L. Dyer, and Vidya Muthukumar
  JMLR 2024
  [arxiv][code]