Research Experience

Predictive Lift modelling : Predicting incremental gains [report, poster]

Course Project in CS229: Machine Learning (September - December 2018)
Designed a predictive response model using uplift modelling technique to predict the “incremental” effect of an ad campaign on consumer behavior and use it to target the optimal set of customers.

Face Swapping and Harmonization using neural nets [report, poster]

Course Project in CS231N: Convolutional Neural Networks for Visual Recognition (April - June 2018)
Developed an approach for face anonymization via. face detection, anonymization and blending. Trained a 9 layer deep CNN on LFW (Labeled Faces in the Wild) dataset.

Network Analysis of Weighted Signed Bitcoin Network [report, poster]

Studied trust in signed weighted bitcoin network and designed a linear regression based algorithm for trust prediction using social imbalance theory and graph topology.

Parameterized Algorithms for Cycle Hitting Problems [report]

Master's Thesis project under the guidance of Prof. S K Mehta (December 2013 - July 2015)
Devised an deterministic FPT algorithm for Even Cycle Transversal Problem. Better than the currently best known FPT algorithm.

Concentration Bounds for Absolutely Normal Real Numbers [report]

Bachelors Project under guidance of Prof. Satyadev Nandakumar (July - November 2013)
Turing gave a computable construction for Absolutely Normal Real Numbers where he used a concentration bound inequality without proving it. In this paper, we prove a slightly weaker version of the inequality and show that Turing's construction still goes through. Paper to be submitted at Information Processing Letters.

Game Arguments in Algorithm Information Theory [report]

Course Project in CS397 under guidance of Prof. Satyadev Nandakumar (January - April 2013)
Devised an alternate proof for Levin's Coding Theorem using ideas from Game Theory. In addition, we also investigated certain other results of AIT like Friedburg's unique numbering, Muchnik Vyugin Theorem, etc. and at their proof involving Game Theory arguments and tried discovering their proof using Game Theoretical arguments.

Movie Recommender System [report]

Course Project in CS771 under guidance of Prof. Harish Karnick (July - November 2013)
Devised a recommender system to guess movie ratings given by a user using an improved version of Matrix Factorization algorithm for movie recommendation used in Netflix contest. Algorithm used is a hybrid filtering algorithm which uses both movie's and user's attributes.

Hand Gesture Recognition using Microsoft's Kinect [report]

Course Project in CS365 under guidance of Prof. Amitabha Mukerjee (March - April 2012)
Recognized robust hand gestures by applying FEMD (Field Earths Mover Distance) on shape of hand extracted by Kinect. Hacked Kinect to detect robust hand gestures using PointCloud Library and OpenCV. Accuracy of nearly 75% reported.

Hypercomputation: Beyond Turing Machines [report]

Course Project in PHI455 under guidance of Prof. A V R Sarma (July - November 2013)
Studied various proposals of extension of Turing Machines and their feasibility. Looked at a few plausible possibilities of hypercomputation and devised a theoretical model for the same.