Naman Agarwal

Email: naman33k@gmail.com

I am a Senior Research Scientist at Google AI Princeton. I graduated with a PhD in Computer Science from Princeton University where I was fortunate to be advised by Elad Hazan. Previously, I graduated with a Masters of Science in Computer Science at the University of Illinois Urbana-Champaign. I was advised at UIUC by Prof. Alexandra Kolla. I did my undergraduate at the Computer Science and Engineering Department at IIT Bombay in 2012. Here I was advised by Prof. Abhiram Ranade.

CV  /  Google Scholar  /  Github  /  LinkedIn

Since my PhD I have been interested in Optimization for Machine Learning with a focus on faster optimization methods for Deep Learning. Recently I have also been working on optimization methods for Robust Online Control and Reinforcement Learning. I am also interested in privacy and optimization aspects of Federated Learning.

• Google AI Blogpost for our work on Mechanical Ventilation is now live
• Paper on Reverse Experience Replay for RL to appear in ICLR 2022.
• We recently resolved the open problem of obtaining a fast algorithm for Online Multiclass Logistic Regression. Paper to be presented at ALT 2022. Preprint here.
• Paper on the Skellam Mechanism for Differential Privacy accepted to Neurips 2021. The paper is also accepted for a Contributed Talk at the Privacy Preserving Machine Learning Workshop 2021. Preprint coming soon.
• Papers on Non-Stochastic Conrol for Iterated Learning Control and Fractal Learning Rates presented at ICML 2021.
• Our work Machine Learning for Mechanical Ventilation Control featured in Princeton Press. Preprint here.

Learning Rate Grafting: Transferability of Optimizer Tuning
Naman Agarwal, Rohan Anil, Elad Hazan, Tomer Koren and Cyril Zhang
Preprint - Open Review Submission

Online Target Q-learning with Reverse Experience Replay: Efficiently finding the Optimal Policy for Linear MDPs
Naman Agarwal, Syomantak Chaudhuri, Prateek Jain, Dheeraj Mysore Nagaraj and Praneeth Netrapalli
International Conference for Learning Representations(ICLR), 2022
Arxiv

Efficient Methods for Online Multiclass Logistic Regression
Naman Agarwal, Satyen Kale and Julian Zimmert
International Conference on Algorithmic Learning Theory(ALT), 2022
Arxiv

The Skellam Mechanism for Differentially Private Federated Learning
Naman Agarwal, Peter Kairouz and Ken Ziyu Liu
Conference on Neural Information Processing Systems(NeurIPS), 2021
Arxiv

Machine Learning for Mechanical Ventilation Control
Daniel Suo, Cyril Zhang, Paula Gradu, Udaya Ghai , Xinyi Chen, Edgar Minasyan, Naman Agarwal, Karan Singh, Julienne LaChance, Tom Zajdel, Manuel Schottdorf, Daniel Cohen and Elad Hazan
Extended Abstract - Machine Learning for Health (ML4H), 2021
Arxiv, Code - Github, Press - Princeton

A Regret Minimization Approach to Iterative Learning Control
Naman Agarwal, Elad Hazan, Aniruddha Majumdar and Karan Singh
International Conference on Machine Learning (ICML), 2021
Arxiv

Acceleration via Fractal Learning Rate Schedules
Naman Agarwal, Surbhi Goel and Cyril Zhang
International Conference on Machine Learning (ICML), 2021
Arxiv

Deluca -- A Differentiable Control Library: Environments, Methods, and Benchmarking
Paula Gradu, John Hallman, Daniel Suo, Alex Yu, Naman Agarwal, Udaya Ghai , Karan Singh, Cyril Zhang, Aniruddha Majumdar and Elad Hazan
Neurips 2020 Workshop on Differentiable computer vision. graphics and physics in machine learning
Arxiv, Code - Github

A Deep Conditioning Treatment of Neural Networks
Naman Agarwal, Pranjal Awasthi and Satyen Kale
International Conference on Algorithmic Learning Theory(ALT), 2021
Arxiv

Stochastic Optimization with Laggard Data Pipelines
Naman Agarwal, Rohan Anil, Tomer Koren, Kunal Talwar and Cyril Zhang
Conference on Neural Information Processing Systems(NeurIPS), 2020
Arxiv

Boosting for Dynamical Systems
Naman Agarwal, Nataly Brukhim, Elad Hazan and Zhou Lu
International Conference on Machine Learning (ICML), 2020
Arxiv

Adaptive regularization with cubics on manifolds
Naman Agarwal, Nicolas Boumal, Brian Bullins and Coralia Cartis
Mathematical Programming, 2020
Arxiv

Extreme Tensoring for Low-Memory Preconditioning
Xinyi Chen, Naman Agarwal, Elad Hazan, Cyril Zhang and Yi Zhang
International Conference for Learning Representations, (ICLR) 2020
Arxiv

Leverage Score Sampling for Faster Accelerated Regression and ERM
Naman Agarwal, Sham Kakade, Rahul Kidambi, Praneeth Nethrapalli, Aaron Sidford and Yin Tat-Lee
Conference on Algorithmic Learning Theory(ALT) 2020
Arxiv

Logarithmic Regret for Online Control
Naman Agarwal, Elad Hazan and Karan Singh
Conference on Neural Information Processing Systems(NeurIPS) 2019, Oral Presentation
Arxiv

Learning in Non-convex Games with an Optimization Oracle
Alon Gonen, Naman Agarwal and Elad Hazan
Conference on Learning Theory(COLT), 2019
Arxiv

Online Control with Adversarial Disturbances
Naman Agarwal, Brian Bullins, Elad Hazan, Sham Kakade and Karan Singh
International Conference on Machine Learning (ICML), 2019
Arxiv

The Case for Full-Matrix Adaptive Regularization
Naman Agarwal, Brian Bullins, Xinyi Chen, Elad Hazan, Karan Singh, Cyril Zhang and Yi Zhang
International Conference on Machine Learning (ICML), 2019
Arxiv

cpSGD: Communication-efficient and differentially-private distributed SGD
Naman Agarwal, Ananda Theertha Suresh, Felix Yu , Sanjiv Kumar and H. Brendan Mcmahan
Conference on Neural Information Processing Systems(NeurIPS) 2018, Spotlight
Arxiv

Lower Bounds for Higher-Order Convex Optimization
Naman Agarwal, Elad Hazan
Conference on Learning Theory(COLT), 2018
Arxiv

The Price of Differential Privacy For Online Learning
Naman Agarwal and Karan Singh
International Conference on Machine Learning (ICML), 2017
Arxiv

Finding Approximate Local Minima for Nonconvex Optimization in Linear Time
Naman Agarwal, Zeyuan Allen-Zhu, Brian Bullins, Elad Hazan and Tengyu Ma
Symposium on Theory of Computing (STOC), 2017
Arxiv

Second Order Stochastic Optimization in Linear Time
Naman Agarwal, Brian Bullins and Elad Hazan
Journal of Machine Learning Research (JMLR), 2017
Preliminary results presented at the Optimization Methods for the Next Generation of Machine Learning workshop at ICML 2016
Arxiv/ Code/ Poster

Multisection in the Stochastic Block Model using Semidefinite Programming
Naman Agarwal, Afonso S. Bandeira, Konstantinos Koiliaris and Alexandra Kolla
Compressed Sensing and Its Applications: Second International MATHEON Conference - 2015
Arxiv

On the Expansion of Group-based Lifts
Naman Agarwal, Karthekeyan Chandrasekaran, Alexandra Kolla and Vivek Madan
SIAM Journal on Discrete Mathematics, Volume 33, Issue 3
21st International Workshop on Randomization and Computation (RANDOM) - 2017
Arxiv

Unique Games on the Hypercube
Naman Agarwal, Guy Kindler, Alexandra Kolla and Luca Trevisan
Chicago Journal of Theoretical Computer Science - 2014
Link

The template was borrowed from this and hence this (which is awesome)