Research Scholar
Industrial Engineering and Operations Research
Indian Institute of Technology Bombay
Supervisor Prof. N Hemchandra
Research Interests
Loss functions in binary classification problems, Label noise models, Statistical learning theory, Explaianability of learning models, Game theory, Queuing theory, Bioinformatics for big data, Computational neuroscience.
Research Summary
For classification problems, in the current scenario, where crowdsourcing is the in thing, there is a high chance of the labels being incorrect (flipped) due to lack of expertise or annotation error, etc. We are interested in devising cost sensitive (misclassification cost is class dependent) algorithms that are tolerant to label noise. Also, we are interested in understanding the feature-feature and feature-label interaction from a cooperative game theory perspective.
Conferences & Workshops
- Presented joint work with Prof N. Hemachandra, titled " Cost sensitive learning in the presence of symmetric label noise " in April 2019 at PAKDD 2019, Macau, China Accepted Paper.
- Presented a poster titled "GANs based label noise robust classification" at the Neuro-inspired computing course held at IRCN, Tokyo, Japan in March 2019.
- Presented joint work with Prof N. Hemachandra, titled " Scalable linear classifiers based on exponential loss functions" in January 2018 at CoDS-COMAD 2018, Goa, India Accepted Paper.
- Presented joint work with Prof N. Hemachandra, titled " Shapley value based binary classifiers" in December 2017 at WINE 2017, Bengaluru, India.
- Presented joint work with Prof N. Hemachandra and Kishore Patil, titled " Equilibrium sets of some GI/M/1 queues " in July 2016 at ECQT 2016, Toulouse, France.
- Attended Machine Learning Summer School, 2015 held in Kyoto, Japan.
Communicated Papers & Technical Reports
- Tripathi, S., Hemachandra, N. and Trivedi, P., " On Feature Interactions Identified by Shapley Values of Binary Classification Games " Communicated.
- Hemachandra, N., Tripathi, S. and Patil, K., "Equilibrium sets of some GI/M/1 queues (with more examples)" Technical Report.
Academic Background
- M.Sc-Ph.D. in Operations Research (2012- present), Industrial Engineering and Operations Research, IIT Bombay, Mumbai
- B.Sc (Hons) Statistics (2009-2012) Lady Sri Ram College for Women, Delhi University, Delhi
Other interests
Exploring new places, Trekking, Music, Indian classical dance, Reading fiction, Being a polyglot.
Contact
Email Id: sandhya.tripathi@iitb.ac.in
IEOR lab: (+91) 022 2576 4807