Devavrat Shah

Professor, Department of EECS

Member, Institute for Data, Systems and Society, Laboratory for Information and Decision Systems, Statistics and Data Science Center

Principal Investigator, MIT Institute for Foundations of Data Science

Massachusetts Institute of Technology

Adjunct Professor, Tata Institute of Fundamental Research

Research Interest

My current research focus is Social Data Processing (see here) with a particular interest in practical algorithms for Statistical inference. Over years, I have been interested in distributed network algorithms, stochastic networks and network information theory. The primary challenge that has been driving my research is the need for designing high-performant, extremely simple algorithms under severe resource constraints.

Research Projects

Luck versus Skill in Fantasy Sports [LuckOrSkill]

Time Series Analysis using Matrix Estimation [TimeMatrix]

Crowd-Sourcing meets Matrix Estimation [CrowdMatrix]

Wonders of Nearest Neighbors [Twitter, BitCoin, Medical Imaging, Monograph]

Robust Synthetic Control [SynCMatrix1]

Sparse Matrix Estimation and Iterative Collaborative Filtering [MatrixICF1, MatrixCF2, MonotoneMatrix]

Data Center Scheduling: Theory and Practice [Fastpass, Flowtunes, BaselineOpt]

Blind Regression and Collaborative Filtering [BlindReg1, BlindReg2]

Graphical Model Learning [Hardness, Dynamic, AntiFerro]

Recommendation Systems [Item-ItemCF, User-UserCF]

Rumor Source Detection [RumorCentrality1, RumorCentrality2]

Ranking, Comparisons and Distribution over Permutations [RankCentrality1, RankCentrality2, MixedMNL-Tensor, MixedMNL, RankAgg] 

Crowd-Sourcing [Crowd1, Crowd2, Crowd3, Crowd4]

Computing Discrete Choice from Limited Data [Sparse1, Sparse2]

Gossip Algorithm [Monograph, GossipAvg1, GossipAvg2, GossipSeparable1, GossipSeparable2, GossipDyn, GossipCvx, GossipCoding]

Graphical Models, Belief Propagation and MAP Inference [BPMatching, BPIndSet, BPAssign, BP=LPTightness]

Graphical Models, Belief Propagation and MARG Inference/Log Partition Function [BPFixedPolyTime]

Graphical Model Inference and Local Algorithms [LocalGeom, LocalMinorEx, PartMerge]

Delay Optimal Scheduling and Computation Hardness [IQSwitch, SwitchedNet, Hardness]

Statistics, Modeling for Circuits [ImpSamp1, ImpSamp2, EnergyRedn]

Wireless Medium Access and Distributed Scheduling [GossipSch, GlauberSch1, GlauberSch2, GlauberRateAlln, ReversibleSch]

Network (and) Information Theory [ThputDelay1, ThputDelay2, ThputDelay3, NIT0, NIT1, NIT2, NIT3]

Switch Scheduling: Randomization, Heavy Traffic and Belief Propagation [Randomized, MaxWtDiffusion, LogWtOpt, BeliefPropSchd, MessPassSchSurvey]