Saravanan Thirumuruganathan

Hi! My name is Saravanan Thirumuruganathan (if you cannot figure out how to pronounce my name, dont worry - my nickname is Sara). I work as a scientist in the Data Analytics group at QCRI (Qatar Computing Research Institute).

I completed my PhD in University of Texas at Arlington where I worked with Dr.Gautam Das on diverse problems in hidden databases, sampling of graphs and crowdsourcing. I did my Masters also at University of Texas at Arlington. I worked with Dr.Manfred Huber on building Bayesian Networks from rules.

My research interests are in the broad areas of Data Mining, Machine Learning and Artificial Intelligence. In QCRI, I work on novel problems in Data Cleaning and Big Data Analytics.

Keywords: Data Cleaning, Data Analytics, Data Exploration, Crowdsourcing, Social Networks, Data Mining, Machine Learning, Graph Analytics.

News and Highlights

  • Jul 2018 Hooray! Overall 5 research papers accepted in VLDB 2018!
  • Jun 2018 "Distributed Representations of Tuples for Entity Resolution" accepted in VLDB 2018. We show how to use Deep Learning to tackle the challenging problem of Entity Resolution!
  • Jun 2018 "Efficient Construction of Approximate Ad-Hoc ML models Through Materialization and Reuse" accepted in VLDB 2018. We describe a number of lightweight techniques to construct approximate ML models for exploratory purposes.
  • Jun 2018 "RHEEM: A Cross-Platform Data Processing System" accepted in VLDB 2018. This is a Systems paper (my first!) published under the "Innovative Systems and Applications" track. This describes Rheem, QCRI's cross platform system.
  • Mar 2018 "Leveraging Similarity Joins for Signal Reconstruction" accepted in VLDB 2018. We use some cool tricks from DB such as similarity joins to dramatically speed up a classic problem in Linear Algebra that has diverse applications in networking, image reconstruction etc.
  • Mar 2018 "Discovery of Genuine Functional Dependencies from Relational Data with Missing Values" accepted in VLDB 2018. We discuss the impact on NULL values on the discovery of Functional Dependencies. We propose a measure to quantify the robustness of a FD.
  • Jan 2018 "Robust Road Map Inference through Network Alignment of Trajectories" has been accepted in SDM 2018. We discuss an elegant graph theoretic approach to construct road maps from GPS trajectories.