Research & Publications

My research interest lies at the intersection of machine learning, signal processing, statistics, and information theory and I work on data analysis problems that arise from diverse applications. In particular, my research vision is to develop a rich suite of techniques that take a holistic approach to the integrated design of learning algorithms and data acquisition systems. I have been designing the next generation of machine learning algorithms that leverage latent structures like graphs, subspaces, manifolds, and clusters to use data effectively while taking into account the constraints imposed by the data acquisition process. This often involves the design of pooling strategies for experiments, compressed data acquisition and transmission strategies, and strategies for interactivity of the learning algorithm with both its environment and with human experts.

Preprints

Coalescent-Based Species Tree Estimation: A Stochastic Farris Transform

G. Dasarathy, E. Mossel, R. Nowak, and S. Roch
Submitted, July 2017
arXiv

Multi-fidelity Gaussian Process Bandit Optimisation

K. Kandasamy, G. Dasarathy, J. Oliva, B. Poczos, and J. Schneider
Submitted.
pdf    arXiv

Journal & Journal-Style CS Conference Papers

Multi-fidelity Bayesian Optimisation with Continuous Approximations

K. Kandasamy, G. Dasarathy, B. Poczos, and J. Schneider
(ICML 17) International Conference on Machine Learning, July 2017
arXiv

The Multi-fidelity Multi-armed Bandit

K. Kandasamy, G. Dasarathy, B. Poczos, and J. Schneider
(NIPS 16) Advances in Neural Information Processing Systems, Barcelona, Spain, December 2016

Gaussian Process Bandit Optimization with Multi-fidelity Evaluations

K. Kandasamy, G. Dasarathy, J. Oliva, B. Poczos, and J. Schneider
(NIPS 16) Advances in Neural Information Processing Systems, Barcelona, Spain, December 2016

Active Learning Algorithms for Graphical Model Selection

G. Dasarathy, A. Singh, M. F. Balcan, and J. H. Park
(AISTATS 16) International Conference on Artificial Intelligence and Statistics (AISTATS) 2016, Cadiz, Spain, May 2016 [Full Oral Presentation]
pdf    supplementary   publisher's website

S2: An Efficient Graph Based Active Learning Algorithm with Application to Nonparametric Classification

G. Dasarathy, R. Nowak, and X. Zhu
(COLT 15) Conference on Learning Theory, Paris, France, July 2015
pdf    arXiv   publisher's website

Data Requirement for Phylogenetic Inference from Multiple Loci: A New Distance Method

G. Dasarathy, R. Nowak, and S. Roch
(TCBB) IEEE Transacations on Computational Biology and Bioinformatics, Vol 12, Issue 2, April 2015
pdf    arXiv   publisher's website

Sketching Sparse Matrices, Covariances, and Graphs via Tensor Products

G. Dasarathy, P. Shah, B. N. Bhaskar, and R. Nowak
(IT) IEEE Transactions on Information Theory, Vol 61, Issue 3, March 2015
pdf    arXiv   publisher's website

Efficient Network Tomography for Internet Topology Discovery

B. Eriksson, G. Dasarathy, P. Barford, and R. Nowak
(ToN) IEEE/ACM Transactions on Networking, Vol 20, Issue 3, June 2012
pdf   publisher's website

Active Clustering: Robust and Efficient Hierarchical Clustering using Adaptively Selected Similarities

B.Eriksson, G. Dasarathy, A. Singh, and R. Nowak
(AISTATS 11) International Conference on Artificial Intelligence and Statistics, Ft. Lauderdale, Florida, 2011
pdf    arXiv    project webpage    code

Toward the Practical Use of Network Tomography for Internet Topology Discovery

B.Eriksson, G. Dasarathy, P. Barford, and R. Nowak
(INFOCOM 10) IEEE International Conference on Computer Communications, San Diego, California, March 2010
pdf    publisher's website

Conference & Workshop Papers

DeepCodec: Adaptive Sensing and Recovery via Deep Convolutional Neural Networks

A. Mousavi, G. Dasarathy, and R. Baraniuk
(Allerton 17) 55th Annual Allerton Conference on Communication, Control, and Computing, October 2017
arXiv

Sketched Covariance Testing: A Compression-Statistics Tradeoff

G. Dasarathy, P. Shah, and R. Baraniuk
(ISIT 17) IEEE International Symposium of Information Theory, Aachen, Germany, June 2017

On Computational and Statistical Tradeoffs in Matrix Completion with Graph Information

G. Dasarathy*, N. Rao*, and R. Baraniuk
(SPARS 17) Signal Processing with Adaptive Sparse Structured Representations Workshop, Lisbon, Portugal, June 2017 [Full Oral Presentation]
*authors contributed equally

New Sample Complexity Bounds for Phylogenetic Inference from Multiple Loci

G. Dasarathy, R. Nowak, and S. Roch
(ISIT 14) IEEE International Symposium on Information Theory, Honolulu, Hawaii, July 2014
pdf    publisher's website

Upper and Lower Bounds on the Reliability of Content Identification

G. Dasarathy, and S. Draper
(IZS 14) International Zürich Seminar on Communications, Zürich, Switzerland, February 2014
S. D. invited

Sketching Sparse Covariance Matrices and Graphs

G. Dasarathy, P. Shah, B. Bhaskar, and R. Nowak
(NIPS 13 Workshop) NIPS workshop on Randomized Methods in Machine Learning, December 2013
pdf

Covariance Sketching

G. Dasarathy, P. Shah, B. Bhaskar, and R. Nowak
(Allerton 12) Allerton Conference on Communication and Control, UIUC. October 2012
R. N. invited
pdf    publisher's website

On Reliability of Content Identification from Databases based on Noisy Queries

G. Dasarathy, and S. Draper
(ISIT 11) IEEE International Symposium on Information Theory, St. Petersburg, Russia, July 2011
pdf    publisher's website

Reliability in Noisy Search

G. Dasarathy, and S. Draper
(ITA 11) UCSD Workshop on Information Theory and Applications, San Diego, California, February 2011
S. D. invited

Thesis

Data Efficient and Robust Algorithms for Reconstructing Large Graphs

Advisors: Robert Nowak and Stark Draper
Department of Electrical and Computer Engineering, University of Wisconsin - Madison, August 2014
pdf

Unpublished Notes

A Simple Probability Trick for finding The Expected Maximum of n Random Variables.

This document outlines a simple method for finding the expected value of the maximum of n random variables.
pdf

"Non-Typewise" Method for Sharper Upper Bounds

A method (initially due to Massey) to sharpen upper bounds obtained using the method of types.
pdf