Chirag Gupta

I am now at Bloomberg AI in New York! Please reach out if you’d like to know more about what we do.

Previously, I spent five wonderful years at Carnegie Mellon University, where I obtained a PhD in Machine Learning under the guidance of Aaditya Ramdas. My research was awarded the Bloomberg Data Science Fellowship. My PhD dissertation, titled “Post-hoc calibration without distributional assumptions”, can be found here. Earlier, I was a Research Fellow at Microsoft Research, India with Prateek Jain. I did my undergrad (B. Tech) in Computer Science at IIT Kanpur.

Publications and preprints

(in reverse chronological order of first preprint)

• Parity Calibration
  Youngseog Chung, Aaron Rumack, Chirag Gupta
  UAI 2023 (oral). [arxiv]

• Online Platt Scaling with Calibeating
  Chirag Gupta, Aaditya Ramdas
  ICML 2023. [arxiv]

• Faster online calibration without randomization: interval forecasts and the power of two choices
  Chirag Gupta, Aaditya Ramdas
  COLT 2022. [proc] [arxiv]

• Top-label calibration and multiclass-to-binary reductions
  Chirag Gupta, Aaditya Ramdas
  ICLR 2022. [proc] [arxiv] [code]

• Distribution-free calibration guarantees for histogram binning without sample splitting
  Chirag Gupta, Aaditya Ramdas
  ICML 2021. [arxiv] [proc] [code] [talk (from 37’ to 50’)]

• Distribution-free binary classification: prediction sets, confidence intervals and calibration
  Chirag Gupta*, Aleksandr Podkopaev*, Aaditya Ramdas
  Neurips 2020 (spotlight). [arxiv] [proc] [talk (from 24’ to 38’)]

• Nested conformal prediction and quantile out-of-bag ensemble methods
  Chirag Gupta, Arun K Kuchibhotla, Aaditya Ramdas
  Pattern Recognition (Special Issue on Conformal Prediction) 2022. [journal] [arxiv] [code] [talk (from 3’ to 24’)]

• Path length bounds for gradient descent and flow
  Chirag Gupta, Sivaraman Balakrishnan, Aaditya Ramdas
  Journal of Machine Learning Research (JMLR) 2021. [journal] [arxiv] [blog]

• Support recovery for orthogonal matching pursuit: upper and lower bounds
  Raghav Somani*, Chirag Gupta*, Prateek Jain, Praneeth Netrapalli
  Neurips 2018 (spotlight). [proc]

• Protonn: Compressed and accurate knn for resource-scarce devices
  Chirag Gupta, Arun Sai Suggala, Ankit Goyal, Harsha Vardhan Simhadri, Bhargavi Paranjape, Ashish Kumar, Saurabh Goyal, Raghavendra Udupa, Manik Varma, Prateek Jain
  ICML 2017. [proc] [code]

Invited talks

• Provably calibrating ML classifiers without distributional assumptions (TrustML Young Scientist Seminar, September 2022) [link]
• Recent advances in distribution-free uncertainty quantification (International Seminar on Selective Inference, May 2021) [link]

Academic service

• Reviewer: ICML 2020, ALT 2020, Neurips 2021, ICML 2022, JMLR
• TA: Mathematical and Computational Foundations for ML (CMU), Machine Learning, Ethics, and Society (CMU)

Powered by Jekyll and Minimal Light theme.