Hello There

I am Shounak Datta.
I am a Machine Learning Researcher

with Duke University.

About

More About Me

I am currently a Research Associate with the Electrical and Computer Engineering Dept. of Duke University. I hold a Ph.D. in Computer Science from the Indian Statistical Institute, Kolkata, India, an M.E. in Electronics and Telecommunication Engineering from the Jadavpur University, Kolkata, India, and a B.Tech. in Electronics and Communication Engineering from the West Bengal University of Technology, Kolkata, India. I am currently working as a Postdoctoral Associate with the Electrical and Computer Engineering Dept. of Duke University.


I have been working on machine learning, data mining and stochastic optimization for 7 years now. My current research interests include development of learning algorithms that are resilient to the various irregularities characterizing real-world data, such as class imbalance, presence of small disjuncts within classes, missing feature values, etc. I am also interested in multi-objective optimization based formulations of traditional machine learning tasks to overcome the shortcomings imposed by the single-objective optimization frameworks.

I have supervised 20+ undergraduate and postgraduate research interns (from various notable academic institutes such as ISI, Jadavpur University, IIT Guwahati, IIIT Guwahati, IIEST, etc.) on projects leading to papers, technical reports and dissertations since 2013.

Current Research Areas

Machine Learning
  • Deep Learning
  • Class Imbalance
  • Missing Features
  • Natural Language Processing
  • Ambiguity Resolution
  • Support Vector Machines
  • k-Nearest Neighbors
  • Clustering
Optimization
  • Evolutionary Computation
  • Multi-objective Optimization

Selected Research Projects

2013 - Present

Learning Systems resilient to Data Irregularities

Indian Statistical Institute

Data irregularities such as class imbalance, presence of small disjuncts, missing features, etc. are common in real-world data. The aim of this project is to devise learning systems which can effectively handle such data irregularities without incurring severe computational overheads.

2011 - 2013

Analysis and Predicting of Stock Market Indices

Jadavpur University

Stock market indices are seemingly random as they are affected by a large number of variables. However, quantitation analysis has sometimes proven to be effective in predicting the behavior of stock market indices. The aim of this prject was to explore the usefulness of soft-computing techniques in predicting the behavior of stock market indices.

2015 - Present

Automatic determination of the level of overlap between clusters

Indian Statistical Institute

Most real-world datasets are characterized by overlap between the clusters found in the data. However, it is difficult to determine the extent of overlap between clusters of unlabeled data. We aim to devise partitional clustering methods capable of automatically detecting the level of overlap between the clusters.

2011 - 2013

Electrooculogram based Control of Wheelchair for Rehabilitation

Jadavpur University

Human-computer interfaces that can be controlled by eye movements may be very useful for the rehabilitation of severely disabled individuals. Electrooculogram, the bio-potential produced due to eye ball motion can be used to track eye movements. The aim of this project was to devise an effective systems for controlling the movement of wheelchairs using electrooculogram.

15

Journal Articles

5

Conference Articles

3

Preprints/Drafts

3

Lectures

Publications

See My Latest Publications.

The following is a list of my research articles including 15 journal- and 5 peer-reviewed conference-papers. For more details visit my Google Scholar page.

Journal Articles

Conference Articles

Contact

Say Hello.

Drop me an email.

Email

shounak(dot)jaduniv[at]gmail(dot)com