Pratik Patil |
Research | Software | Talks | Personal |
"It is not knowledge, but the act of learning, not possession but the act of getting there, which grants the greatest enjoyment. When I have clarified and exhausted a subject, I turn away from it, in order to go into darkness again." — Carl Friedrich Gauss, Letter to Farkas Bolyai, September 1808 [source: Quotes by Gauss]
I am fortunate to work with a diverse set of collaborators across various disciplines. Each of these disciplines follows different conventions for deciding on authorship order. However, these are just that—conventions. The true value of each contribution, however you define it, is often difficult to quantify. In that spirit, think of every project below as a result of a "team" effort, where every "player" brings their strengths to the field.
What I love the most about collaborations is the sense of intellectual closeness it brings. When you discuss half-baked ideas—often at a level just above talking to yourself—with collaborators, it almost feels like your brain has an extension. Much like how you can say a thousand words to your loved ones without saying anything, I find that you can say (at least) a hundred words to your close collaborators without saying anything, which feels magical to me!
Confidence Intervals for Functionals in Constrained Inverse Problems via Data-adaptive Sampling-based Calibration
Mike Stanley, Pau Batlle, Pratik Patil, Houman Owhadi, and Mikael Kuusela
[paper]
[slides]
Revisiting Model Optimism and Model Complexity in the Wake of Overparameterized Learning
Pratik Patil, Jin-Hong Du, and Ryan Tibshirani
[paper]
[code]
[slides]
Precise Asymptotics of Subagging of Regularized M-estimators
Takuya Koriyama*, Pratik Patil*, Jin-Hong Du, Kai Tan, and Pierre Bellec
[paper]
[code]
[slides]
Optimization-based Frequentist Confidence Intervals for Functionals in Constrained Inverse Problems: Resolving the Burrus Conjecture
Pau Batlle, Pratik Patil, Michael Stanley, Houman Owhadi, and Mikael Kuusela
[paper]
[slides]
[poster]
Mitigating Multiple Descents: A Model-Agnostic Framework for Risk Monotonization {♥}
Pratik Patil, Arun Kuchibhotla, Yuting Wei, and Alessandro Rinaldo
[paper]
[slides]
Implicit Regularization Paths of Weighted Neural Representations
Jin-Hong Du and Pratik Patil
Neural Information Processing Systems (NeurIPS), 2024 [poster]
[paper]
[code]
[openreview]
Corrected Generalized Cross-Validation for a Finite Ensemble of Penalized Estimators
Pierre Bellec†, Jin-Hong Du†, Takuya Koriyama*†, Pratik Patil*†, and Kai Tan†
Journal of the Royal Statistical Society: Series B (JRSSB), 2024
[paper]
[code]
[poster]
[slides]
A Framework for Efficient Model Evaluation through Stratification, Sampling, and Estimation
Riccardo Fogliato, Pratik Patil, Mathew Monfort, and Pietro Perona
European Conference on Computer Vision (ECCV), 2024 [poster]
[paper]
[code]
[slides]
Optimal Ridge Regularization for Out-of-Distribution Prediction
Pratik Patil, Jin-Hong Du, and Ryan Tibshirani
International Conference on Machine Learning (ICML), 2024 [spotlight]
[paper]
[code]
[poster]
[slides]
[openreview]
Confidence Intervals for Error Rates in 1:1 Matching Tasks: Critical Statistical Analysis and Recommendations
Riccardo Fogliato*, Pratik Patil*, and Pietro Perona
International Journal of Computer Vision (IJCV), 2024
[paper]
[code]
[slides]
Failures and Successes of Cross-Validation for Early-Stopped Gradient Descent in High-Dimensional Least Squares
Pratik Patil, Yuchen Wu, and Ryan Tibshirani
International Conference on Artificial Intelligence and Statistics (AISTATS), 2024 [oral]
[paper]
[code]
[poster]
[slides]
Asymptotically Free Ridge Ensembles: Risks, Cross-Validation, and Tuning
Pratik Patil and Daniel LeJeune
International Conference on Learning Representations (ICLR), 2024 [spotlight]
[paper]
[code]
[poster]
[slides]
[openreview]
Generalized Equivalences between Subsampling and Ridge Regularization {♥}
Pratik Patil and Jin-Hong Du
Neural Information Processing Systems (NeurIPS), 2023 [poster]
[paper]
[code]
[poster]
[slides]
[openreview]
Asymptotics of the Sketched Pseudoinverse
Daniel LeJeune*, Pratik Patil*, Hamid Javadi, Richard Baraniuk, and Ryan Tibshirani
SIAM Journal on Mathematics of Data Science (SIMODS), 2023
[paper]
[code]
[poster]
[slides]
Bagging in Overparameterized Learning: Risk Characterization and Risk Monotonization
Pratik Patil*, Jin-Hong Du*, and Arun Kuchibhotla
Journal of Machine Learning Research (JMLR), 2023
[paper]
[code]
[slides]
Subsample Ridge Ensembles: Equivalences and Generalized Cross-Validation
Jin-Hong Du*, Pratik Patil*, and Arun Kuchibhotla
International Conference on Machine Learning (ICML), 2023 [oral]
[paper]
[code]
[poster]
[slides]
[openreview]
Extrapolated Cross-Validation for Randomized Ensembles
Jin-Hong Du, Pratik Patil, Kathryn Roeder, and Arun Kuchibhotla
Journal of Computational and Graphical Statistics (JCGS), 2023
[paper]
[code]
[slides]
Estimating Functionals of the Out-of-Sample Error Distribution in High-Dimensional Ridge Regression {♥}
Pratik Patil, Alessandro Rinaldo, and Ryan Tibshirani
International Conference on Artificial Intelligence and Statistics (AISTATS), 2022 [oral]
[paper]
[code]
[poster]
[slides]
[handout]
Uncertainty Quantification for Wide-Bin Unfolding: One-at-a-Time Strict Bounds and Prior-Optimized Confidence Intervals
Michael Stanley, Pratik Patil, and Mikael Kuusela
IOP Journal of Instrumentation (JINST), 2022
[paper]
[code]
[slides]
Objective Frequentist Uncertainty Quantification for Atmospheric CO2 Retrievals
Pratik Patil, Mikael Kuusela, and Jon Hobbs
SIAM-ASA Journal on Uncertainty Quantification (JUQ), 2022
[paper]
[code]
[code]
[slides]
Uniform Consistency of Cross-Validation Estimators for High-Dimensional Ridge Regression
Pratik Patil, Yuting Wei, Alessandro Rinaldo, and Ryan Tibshirani
International Conference on Artificial Intelligence and Statistics (AISTATS), 2021 [oral]
[paper]
[code]
[poster]
[slides]
[handout]
Uplink-Downlink Duality Between Multiple-Access and Broadcast Channels with Compressing Relays {♥}
Liang Liu†, Ya-Feng Liu†, Pratik Patil†, and Wei Yu†
IEEE Transactions on Information Theory (TIT), 2021
[paper]
[code]
[slides]
An Open Repository of Real-Time COVID-19 Indicators
Alex Reinhart, et al., Lester Mackey, et al., Pratik Patil, et al., Samyak Rajanala, et al., and Ryan Tibshirani
(listing only my local teammates explicitly, among a global team of 67 authors)
Proceedings of the National Academy of Sciences (PNAS), 2021
[paper]
[code]
[api]
[Python client]
[R client]
Generalized Compression Strategy for the Downlink Cloud Radio Access Network
Pratik Patil and Wei Yu
IEEE Transactions on Information Theory (TIT), 2019
[paper]
[code]
[abridged]
Channel Diagonalization for Cloud Radio Access
Liang Liu, Pratik Patil, and Wei Yu
IEEE Wireless Communications Letters (WCL), 2018
[paper]
Hybrid Data-Sharing and Compression Strategy for Downlink Cloud Radio Access Network
Pratik Patil, Binbin Dai, and Wei Yu
IEEE Transactions on Communications (TCOM), 2018
[paper]
[code]
[poster]
Layered Construction for Low-Delay Streaming Codes
Ahmed Badr, Pratik Patil, Ashish Khisti, Wai-Tian Tan, and John Apostolopoulos
IEEE Transactions on Information Theory (TIT), 2016
[paper]
[code]
[poster]
Precise Model Benchmarking with Only a Few Observations
Riccardo Fogliato, Pratik Patil, Nil-Jana Akpinar, and Mathew Monfort
Empirical Methods in Natural Language Processing (EMNLP), 2024
[paper]
[code]
[openreview]
Revisiting Model Complexity in the Wake of Overparameterized Learning
Pratik Patil and Ryan Tibshirani
Theory of Overparameterized Learning (TOPML), 2022 [spotlight]
[paper]
[slides]
[handout]
Mitigating Multiple Descents: Model-Agnostic Risk Monotonization in High-Dimensional Learning
Pratik Patil, Arun Kuchibhotla, Yuting Wei, and Alessandro Rinaldo
Theory of Overparameterized Learning (TOPML), 2021 [lightning]
[paper]
[slides]
An Uplink-Downlink Duality for Cloud Radio Access Network
Liang Liu, Pratik Patil, and Wei Yu
International Symposium on Information Theory (ISIT), 2016
[paper]
[slides]
Performance Comparison of Data-Sharing and Compression Strategies for Cloud Radio-Access Networks
Pratik Patil, Binbin Dai, and Wei Yu
European Signal Processing Conference (EUSIPCO), 2015 [invited]
[paper]
[code]
[slides]
Hybrid Compression and Message-Sharing Strategy for the Downlink Cloud Radio-Access Network
Pratik Patil and Wei Yu
Information Theory and Applications Workshop (ITA), 2014 [invited]
[paper]
[code]
[slides]
Network Coding Design for Multi-Source Multi-Relay Cooperative Wireless Networks
Pratik Patil, Wasif Khan, Ratnajit Bhattacharjee, and Sanjay Bose
IEEE Region Ten Conference (Tencon), 2013
[paper]
[slides]
Delay-Optimal Streaming Codes under Source-Channel Rate Mismatch {♥}
Pratik Patil, Ahmed Badr, Ashish Khisti, and Wai-Tian Tan
Asilomar Conference on Signals, Systems and Computers (Asilomar), 2013 [best student paper award]
(Winning this award for one of my first undergraduate projects was what set me on the research path and remains one of my favorite dearest papers!
Many thanks to Ashish for carefully abstracting out a nice puzzle (check out the puzzle in the slides) and for believing that a naive undergrad can play with it!)
[paper]
[poster]
[slides]
[handout]
Streaming Erasure Codes under Mismatched Source-Channel Frame Rates
Pratik Patil, Ahmed Badr, and Ashish Khisti
Canadian Workshop on Information Theory (CWIT), 2013
[paper]
[slides]
[handout]
Downlink Transmission Strategies for Cloud Radio-Access Networks
Pratik Patil
Department of Electrical and Computer Engineering, University of Toronto, 2016
[thesis]
[slides]
[handout]
Facets of Regularization in High-Dimensional Learning: Cross-Validation, Risk Monotonization, and Model Complexity
Pratik Patil
Departments of Statistics and Machine Learning, Carnegie Mellon University, 2022 [best thesis award nomination]
[thesis]
[slides]
[handout]