Aaditya K. Ramdas           
      Tenure-track Assistant Professor           
      Department of Statistics and Data Science           
      (affl.) Machine Learning Department           
      Carnegie Mellon University (CMU)            

Bio: I finished my PhD in 2015 in the Machine Learning Department and the Department of Statistics at Carnegie Mellon University (under Larry Wasserman and Aarti Singh). Following this, I did a 3-year postdoc in the departments of EECS and Statistics at UC Berkeley (under Martin Wainwright and Michael Jordan).

Keywords: scientific reproducibility, false discovery rate, adaptive data analysis, selective inference, hypothesis testing, online experimentation, interactive learning, convex optimization, multi-armed bandits, sequential analysis, nonparametric inference.

Interests: My research spans theory and algorithms in machine learning and statistical inference, with applications to science and technology. One line of recent work focuses on the theme of reproducibility in science and technology (multiple hypothesis testing, adaptive data analysis, selective inference), by designing new algorithms for controlling false discoveries in novel static and dynamic settings. Another line of work involves active sequential experimentation (interactive learning, multi-armed bandits, martingale concentration, confidence sequences), by designing new algorithms that work in online or streaming data settings. I also work on problems in convex optimization and nonparametric statistics.

Curriculum Vitae (pdf)

I only list, by topic, rigorously peer-reviewed, full-length papers at reputable journals or conferences.
I exclude all short workshop/conference papers, poster/talk abstracts and papers with little/no review.
* indicates an equally contributing (often student) author.

False Discovery Rate control (in structured, dynamic, interactive settings)

Towards "simultaneous selective inference": post-hoc bounds on the FDP
E. Katsevich, A. Ramdas [arxiv]
(in submission, the Annals of Statistics)
A unified treatment of multiple testing with prior knowledge using the p-filter
A. Ramdas, R. F. Barber, M. Wainwright, M. Jordan [arxiv] [code]
(in revision, the Annals of Statistics)
P-filter: multi-layer FDR control for grouped hypotheses
R. F. Barber*, A. Ramdas* [arxiv] [code] [JRSSB]
(JRSSB) Journal of the Royal Statistical Society -- Series B (Methodology), 2016
Optimal rates and tradeoffs for multiple testing
M. Rabinovich, A. Ramdas, M. Wainwright, M. Jordan [arxiv]
(SS) Statistica Sinica, 2018
MAB-FDR: Multi (A)rmed/(B)andit testing with online FDR control
F. Yang, A. Ramdas, K. Jamieson, M. Wainwright [arxiv] [code] [30-min talk] [NIPS] [spotlight talk]
(NIPS) 31st Conference on Neural Information Processing Systems, Long Beach, 2017
Online control of the false discovery rate with decaying memory
A. Ramdas, F. Yang, M. Wainwright, M. Jordan [arxiv] [3-min summary] [NIPS] [15-min talk from 44:00] [oral]
(NIPS) 31st Conference on Neural Information Processing Systems, Long Beach, 2017
QuTE: decentralized FDR control on sensor networks
A. Ramdas, J. Chen, M. Wainwright, M. Jordan [code] [CDC]
(CDC) IEEE Conference on Decision and Control, Melbourne, 2017
DAGGER: A sequential algorithm for FDR control on DAGs
A. Ramdas, J. Chen, M. Wainwright, M. Jordan [arxiv] [code]
(BM) Biometrika, 2018
SAFFRON: an adaptive algorithm for online FDR control
A. Ramdas, T. Zrnic, M. Wainwright, M. Jordan [arxiv]
(ICML) 35th Intl. Conference on Machine Learning, Stockholm, 2018
STAR: A general interactive framework for FDR control under structural constraints
L. Lei, A. Ramdas, W. Fithian [arxiv] [movies]
(in submission, Journal of the American Statistical Association)

Hypothesis testing (in nonparametric, structured or high-dimensional settings)

Generative models and model criticism via optimized Maximum Mean Discrepancy
D. Sutherland, H. Tung, H. Strathmann, S. De, A. Ramdas, A. Smola, A. Gretton [arxiv] [ICLR] [poster] [code]
(ICLR) 5th International Conference on Learning Representations, Toulon, 2017
Adaptivity & computation-statistics tradeoffs for kernel & distance based high-dimensional two sample testing
A. Ramdas, S. Reddi, B. Poczos, A. Singh, L. Wasserman [arxiv]
(in revision)
Minimax lower bounds for linear independence testing
D. Isenberg*, A. Ramdas*, A. Singh, L. Wasserman [arxiv][ISIT]
(ISIT) IEEE International Symposium on Information Theory, Barcelona, 2016
Fast two-sample testing with analytic representations of probability measures
K. Chwialkowski, A. Ramdas, D. Sejdinovic, A. Gretton [arxiv] [github] [NIPS]
(NIPS) 29th Conference on Neural Information Processing Systems, Montreal, 2015
Nonparametric independence testing for small sample sizes
A. Ramdas*, L. Wehbe* [arxiv] [IJCAI] [20-min oral]
(IJCAI) 24th International Joint Conference on Artificial Intelligence, Buenos Aires, 2015
On the high-dimensional power of a linear-time two sample test under mean-shift alternatives
S. Reddi*, A. Ramdas*, A. Singh, B. Poczos, L. Wasserman [AISTATS] [arxiv] [pdf] [supp] [errata]
(AISTATS) 18th International Conference on Artificial Intelligence and Statistics, San Diego, 2015
On the decreasing power of kernel and distance based nonparametric hypothesis tests in high dimensions
A. Ramdas*, S. Reddi*, B. Poczos, A. Singh, L. Wasserman [AAAI] [arxiv] [pdf] [supp]
(AAAI) 29th AAAI Conference on Artifical Intelligence, Austin, 2015
Classification accuracy as a proxy for two sample testing
A. Ramdas, A. Singh, L. Wasserman [arxiv]
(in revision)
On Wasserstein two sample testing and related families of nonparametric tests
A. Ramdas*, N. Garcia*, M. Cuturi [arxiv] [Entropy]
(Ent) Entropy, Special Issue on Statistical Significance and the Logic of Hypothesis Testing, 2017

Convex optimization

Iterative methods for solving factorized linear systems
A. Ma, D. Needell, A. Ramdas [arxiv] [SIMAX]
(SIMAX) SIAM Journal on Matrix Analysis and Applications, 2017
Rows vs columns : randomized Kaczmarz or Gauss-Seidel for ridge regression
A. Hefny*, D. Needell*, A. Ramdas*, [arxiv]
(SISC) SIAM Journal on Scientific Computing, 2017
Towards a deeper geometric, analytic and algorithmic understanding of margins
A. Ramdas, J. Pena [arxiv] [OMS]
(OMS) Optimization Methods and Software, 2015
Convergence properties of the randomized extended Gauss-Seidel and Kaczmarz methods
A. Ma*, D. Needell*, A. Ramdas* [arxiv] [SIMAX]
(SIMAX) SIAM Journal on Matrix Analysis and Applications, 2015
Fast & flexible ADMM algorithms for trend filtering
A. Ramdas*, R. Tibshirani* [arxiv] [JCGS] [github `glmgen'] [50 min. talk]
(JCGS) Journal of Computational and Graphical Statistics, 2015
Margins, kernels and non-linear smoothed perceptrons
A. Ramdas, J. Pena [arxiv] [ICML] [pdf] [supp] [20-min oral]
(ICML) 31st International Conference on Machine Learning, Beijing, 2014
Optimal rates for stochastic convex optimization under Tsybakov's noise condition
A. Ramdas, A. Singh [ICML] [arxiv] [pdf] [supp] [20-min oral]
(ICML) 30th International Conference on Machine Learning, Atlanta, 2013

Other problems of a sequential nature

On the power of online thinning in reducing discrepancy
R. Dwivedi, O. N. Feldheim, Ori Gurel-Gurevich, A. Ramdas [arxiv]
(PTRF) Probability Theory and Related Fields, 2018
Uniform, nonparametric, nonasymptotic confidence sequences
S. Howard, A. Ramdas, J. Sekhon, J. McAuliffe [pre.]
(in preparation)
Function-specific mixing times and concentration away from equilibrium
M. Rabinovich, A. Ramdas, M. Wainwright, M. Jordan [arxiv]
(in revision, Bayesian Analysis)
Sequential nonparametric testing with the law of the iterated logarithm
A. Balsubramani*, A. Ramdas* [arxiv] [UAI]
(UAI) 32nd Conference on Uncertainty in Artificial Intelligence, New York, 2016
An analysis of active learning with uniform feature noise
A. Ramdas, A. Singh, L. Wasserman, B. Poczos [arxiv] [AISTATS] [pdf] [supp] [25-min oral]
(AISTATS) 17th International Conference on Artificial Intelligence and Statistics, Reykjavik, 2014
Algorithmic connections between active learning and stochastic convex optimization
A. Ramdas, A. Singh [arxiv] [ALT] [pdf] [25-min oral]
(ALT) 24th International Conference on Algorithmic Learning Theory, Singapore, 2013

Other collaborations

On kernel methods for covariates that are rankings
H. Mania, A. Ramdas, M. Wainwright, M. Jordan, B. Recht [arxiv]
(EJS) Electronic Journal of Statistics, 2018
Decoding from pooled data (II): sharp information-theoretic bounds
A. El-Alaoui, A. Ramdas, F. Krzakala, L. Zdeborova, M. Jordan [arxiv]
(in submission, SIAM Journal on the Mathematics of Data Science)
Decoding from pooled data (I): phase transitions of message passing
A. El-Alaoui, A. Ramdas, F. Krzakala, L. Zdeborova, M. Jordan [arxiv]
(ITIT) IEEE Transactions on Information Theory, 2018
(ISIT) IEEE International Symposium on Information Theory, Aachen, 2017 (short version)
Asymptotic behavior of Lq-based Laplacian regularization in semi-supervised learning
A. El-Alaoui, X. Cheng, A. Ramdas, M. Wainwright, M. Jordan [arxiv][COLT]
(COLT) 29th International Conference on Learning Theory, New York, 2016
Regularized brain reading with shrinkage and smoothing
L. Wehbe, A. Ramdas, R. Steorts, C. Shalizi [arxiv] [AoAS]
(AoAS) Annals of Applied Statistics, 2015
Simultaneously uncovering the patterns of brain regions involved in different story reading subprocesses
L. Wehbe, B. Murphy, P. Talukdar, A. Fyshe, A. Ramdas, T. Mitchell [website] [PLOS] [pdf] [supp]
(PLoS ONE) Public Library of Science ONE, 2014

Theses and Reports

Computational and Statistical Advances in Testing and Learning (PhD thesis)
A. Ramdas [pdf] (Umesh K. Gavaskar Memorial Thesis Award in Statistics)
(CMU) Carnegie Mellon University, Statistics and Machine Learning, 2015
Analysis of burglaries in Pittsburgh (data analysis project)
A. Ramdas [pdf]
(CMU) Carnegie Mellon University, Statistics, 2013
Algorithms for graph similarity and subgraph matching (course project)
D. Koutra, A. Parikh, A. Ramdas, J. Xiang [pdf]
(CMU) Carnegie Mellon University, Machine Learning, 2011
Termination of single-loop linear programs (Bachelor's thesis)
A. Ramdas [pdf]
(IITB) IIT Bombay, Computer Science and Engineering, 2009
Volume-based landmark selection for dimensionality reduction (internship report)
A. Ramdas [pdf]
(INRIA) Geometrica, INRIA Sophia-Antipolis, 2008
Network of timed automata and their symbolic unfolding (internship report)
A. Ramdas [pdf]
(LaBRI) LaBRI, University of Bordeaux I, 2007