Ananda Theertha Suresh Google Research, New York email: theertha at google dot com [Google scholar]

I am a research scientist at Google Research, New York. I obtained my PhD in electrical and computer engineering from University of California, San Diego, where I was advised by Alon Orlitsky. Prior to joining UCSD, I obtained a Bachelor's degree in Engineering Physics from Indian Institute of Technology, Madras.

Research

I am broadly interested in theoretical and algorithmic aspects of machine learning, information theory, and statistics. My current research focus includes differential privacy, federated learning, and domain adaptation.

Publications

• On the benefits of maximum likelihood estimation for Regression and Forecasting, ICLR 2022
  with P. Awasthi, A. Das, R. Sen [pdf]

• Robust Estimation for Random Graphs, Manuscript
  with J. Acharya, A. Jain, G. Kamath, and H. Zhang [pdf]

• Learning with User-Level Privacy, NeurIPS 2021
  with D. Levy, Z. Sun, K. Amin, S. Kale, A. Kulesza, and M. Mohri [pdf]

• Remember what you want to forget: Algorithms for machine unlearning, NeurIPS 2021
  with A. Sekhari, J. Acharya, and G. Kamath [pdf]

• Breaking the centralized barrier for cross-device federated learning, NeurIPS 2021
  with S. Karimireddy, M. Jaggi, S. Kale, M. Mohri, S. Reddi, and S. Stich [pdf]

• Boosting with Multiple Sources, NeurIPS 2021
  with C. Cortes, M. Mohri, and D. Storcheus [pdf]

• FedJAX: Federated learning simulation with JAX, NeurIPS FL workshop 2021
  with J. H. Ro and K. Wu [pdf]

• On the Renyi Differential Privacy of the Shuffle Model, CCS 2021
  with A. Girgis, D. Data, S Diggavi, and P. Kairouz [pdf] (Best paper award)

• Communication-Efficient Agnostic Federated Averaging, Interspeech 2021
  with J. Ro, M. Chen, R. Mathews, and M. Mohri [pdf]

• A discriminative technique for multiple-source adaptation, ICML 2021
  with C. Cortes, M. Mohri, and N. Zhang [pdf]

• Relative deviation margin bounds, ICML 2021
  with C. Cortes and M. Mohri [pdf]

• Wyner-Ziv Estimators: Efficient Distributed Mean Estimation with Side Information, AISTATS 2021
  with P. Mayekar and H. Tyagi [pdf]

• Robust hypothesis testing and distribution estimation in Hellinger distance, AISTATS 2021 [pdf]

• Shuffled Model of Federated Learning: Privacy, Communication and Accuracy Trade-offs, IEEE Journal on Selected Areas in Information Theory
  with A. Girgis, D. Data, S Diggavi, and P. Kairouz [longer version]

• Shuffled Model of Differential Privacy in Federated Learning, AISTATS 2021
  with A. Girgis, D. Data, S Diggavi, and P. Kairouz [pdf]

• A Theory of Multiple-Source Adaptation with Limited Target Labeled Data, AISTATS 2021
  with Y. Mansour, M. Mohri, J. Ro, and K. Wu [pdf]

• HD-cos Networks: Efficient Neural Architectures for Secure Multi-Party Computation, Manuscript
  with W. Jitkrittum, M. Lukasik, F. Yu, and G. Wang [pdf]

• A field guide to federated optimization, Manuscript
  with J. Wang et al [pdf]

• Learning discrete distributions: user vs item-level privacy, NeurIPS 2020
  with Y. Liu, F. Yu, S. Kumar, and M. Riley [pdf]

• Optimal multiclass overfitting by sequence reconstruction from hamming queries, ALT 2020
  with J. Acharya [pdf] (Best paper award)

• Scaffold: Stochastic controlled averaging for federated learning, ICML 2020
  with S. Karimireddy, S. Kale, M. Mohri, S. Reddi, and S. Stich [pdf]

• FedBoost: A Communication-Efficient Algorithm for Federated Learning, ICML 2020
  with J. Hamer and M. Mohri [pdf]

• Three approaches for personalization with applications to federated learning, Manuscript
  with Y. Mansour, M. Mohri, and J. Ro [pdf]

• Can you really backdoor federated learning?, NeurIPS Federated Learning for Data Privacy and Confidentiality workshop 2019
  with Z. Sun, P. Kairouz, and B. McMahan [pdf]

• AdaCliP: Adaptive clipping for private SGD, TPDP workshop 2020
  with V. Pichapati, F. Yu, S Reddi, and S. Kumar [pdf]

• Federated learning of N-gram language models, CoNLL 2019
  with M. Chen, R. Mathews, A. Wong, C. Allauzen, F. Beaufays, and M. Riley [pdf]

• Convergence of Chao unseen species estimator, ISIT 2019
  with N. Rajaraman, P. Chandra, and A. Thangaraj [pdf]

• Approximating probabilistic models as weighted finite automata, Computational Linguistics Journal
  with B. Roark, M. Riley, and V. Schogol [pdf]

• West: Word encoded sequence transducers, ICASSP 2019
  with E. Variani and M. Weintraub [pdf]

• Agnostic federated learning, ICML 2019
  with M. Mohri and G. Sivek [pdf]

• Differentially private anonymized histograms, NeurIPS 2019 [pdf]

• Distilling weighted finite automata from arbitrary probabilistic models, FSMLNP 2019
  with B. Roark, M. Riley, and V. Schogol [pdf]

• Sampled softmax with random fourier features, NeurIPS 2019
  with A. Rawat, J. Chen, F. Yu, and S. Kumar [pdf]

• Advances and Open Problems in Federated learning, Manuscript
  with P. Kairouz et al [pdf]

• Maximum selection and sorting with adversarial comparators, JMLR 2018
  with J. Acharya, M. Falahatgar, A. Jafarpor, and A. Orlitsky [pdf]

• Data amplification: A unified and competitive approach to property estimation, NeurIPS 2018
  with Y. Hao, A. Orlitsky, and Y. Wu [pdf]

• cp-sgd: Communication-efficient and differentially-private distributed SGD, NeurIPS 2018
  with N. Agarwal, F. Yu, S. Kumar, and B. McMahan [pdf] (Spotlight presentation)

• Minimax risk for missing mass estimation, ISIT 2017
  with N. Rajaraman, P. Chandra, and A. Thangaraj [pdf]

• Model-powered conditional independence test, NeurIPS 2017
  with R. Sen, K. Shanmugam, A. Dimakis, and S. Shakkottai [pdf]

• Multiscale quantization for fast similarity search, NeurIPS 2017
  with X. Wu, R. Guo, D. Holtmann-Rice, D. Simcha, F. Yu, and S. Kumar [pdf]

• Lattice rescoring strategies for long short term memory language models in speech recognition, ASRU 2017
  with S. Kumar, M. Nirschl, D. Holtmann-Rice, H. Liao, and F. Yu [pdf]

• Distributed mean estimation with limited communication, ICML 2017
  with F. Yu, H. B. McMahan, and S. Kumar [pdf]

• A unified maximum likelihood approach for optimal distribution property estimation, ICML 2017
  with J. Acharya, H. Das, and A. Orlitsky [pdf] (Best paper award honorable mention)

• Maximum selection and ranking under noisy comparisions, ICML 2017
  with M. Falahatgar, A. Orlitsky, and V. Pichapati [pdf]

• Sample complexity of population recovery, COLT 2017
  with Y. Polyanskiy and Y. Wu [pdf]

• Orthogonal random features, NeurIPS 2016
  with F. Yu, K. Choromanski, D. Holtmann-Rice, and S. Kumar [pdf] (Oral presentation)

• Federated learning: Strategies for improving communication efficiency, NeurIPS PMPML workshop 2016
  with J. Konecny, H. B. McMahan, F. X. Yu, P. Richtarik, and D. Bacon [pdf]

• Optimal prediction of the number of unseen species, PNAS 2016
  with A. Orlitsky and Y. Wu [pdf]

• Learning Markov distributions: Does estimation trump compression?, ISIT 2016
  with M. Falahatgar, A. Orlitsky, and V. Pichapati [pdf]

• Estimating the number of defectives with group testing, ISIT 2016
  with M. Falahatgar, A. Jafarpour, A. Orlitsky, and V. Pichapati [pdf]

• Competitive distribution estimation: Why is Good-Turing good, NeurIPS 2015
  with A. Orlitsky [pdf] [talk] (Best paper award)

• Faster algorithms for testing under conditional sampling, COLT 2015
  with M. Falahatgar, A. Jafarpour, A. Orlitsky, and V. Pichapati [jmlr]

• On learning distributions from their samples, COLT 2015
  with S. Kamath, A. Orlitsky, and V. Pichapati [jmlr]

• Automata and graph compression, ISIT 2015
  with M. Mohri and M. Riley [pdf] [implementation]

• Universal compression of power-law distributions, ISIT 2015
  with M. Falahatgar, A. Jafarpour, A. Orlitsky, and V. Pichapati [pdf]

• Sparse solutions to nonnegative linear systems and applications, AISTATS 2015
  with A. Bhaskara and M. Zaghimoghaddam [arXiv]

• The complexity of estimating Renyi entropy, SODA 2015
  with J. Acharya, A. Orlitsky and H. Tyagi [arXiv]

• Near-optimal-sample estimators for spherical Gaussian mixtures, NeurIPS 2014
  with J. Acharya, A. Jafarpour, and A. Orlitsky [arXiv] [talk at simons]

• Sorting with adversarial comparators and application to density estimation, ISIT 2014
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf]

• Efficient compression of monotone and m-modal distributions, ISIT 2014
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf]

• Poissonization and universal compression of envelope classes, ISIT 2014
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf]

• Sublinear algorithms for outlier detection and generalized closeness testing, ISIT 2014
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf]

• Optimal probability estimation with applications to prediction and classification, COLT 2013
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf] [talk]

• Tight Bounds for Universal Compression of Large Alphabets, ISIT 2013
  with J. Acharya, H. Das, A. Jafarpour, and A. Orlitsky [pdf]

• A competitive test for uniformity of monotone distributions, AISTATS 2013
  with J. Acharya, A. Jafarpour, and A. Orlitsky [pdf]

• Competitive classification and closeness testing, COLT 2012
  with J. Acharya, H. Das, A. Jafarpour, A. Orlitsky, and S. Pan [pdf] [talk]

• On the query computation and verification of functions, ISIT 2012
  with H. Das, A. Jafarpour, A. Orlitsky, and S. Pan [pdf]

• Strong and weak secrecy in wiretap channels, invited paper at ISTC 2010
  A. Subramanian, A. T. Suresh, S. Raj, A. Thangaraj, M. Bloch, and S. W. McLaughlin [pdf]

• Strong secrecy for erasure wiretap channels, ITW 2010
  A. T. Suresh, A. Subramanian, A. Thangaraj, M. Bloch, and S. W. McLaughlin [pdf]

• On optimal timer-based distributed selection For rate-adaptive multi-user diversity systems, NCC 2010
  A. T. Suresh, N. B. Mehta, and V. Shah [pdf] (Best paper award in communications track)