Academics


Resume:
Click here to view my resume

Manuscripts:
  1. Final year thesis on language evolution via inductive learning in communities of artificial agents.
  2. Survey on Information Geometry.
  3. Survey on Tree Alignment methods.
  4. Survey on replicability in science

Undergraduate Research Projects:
I pursued a rather eclectic selection of projects in my undergraduate days. Brief summaries of some of them follow.
  • Statistical models of Language Evolution in Multiple agent communities (pdf) (abstract)
    Advisor: Dr. Harish Karnick, IIT Kanpur
    • Analysis of social interaction and language learning models, for different community graphs, extending a well known syntactic model of language
    • Analyzed language mingling and divergence in communities by stochastically coupling agent based approaches with population dynamics
    • Identified the role of social topologies in shaping the course of linguistic evolution and language divergence
  • Reinforcement Learning based Autoguider for Astrophotography
     (pdf) (abstract)                                                        
    Advisors: Dr. Bernhard Scholkopf and Dr. Jan Peters, Dept of Statistical Inference, Max Planck Institute Tubingen
    • Designed and implemented a two variable Q-learning type formulation for autoguiding a Mach GTO German Equitorial Mount for star-tracking and astrophotography
    • Formulated problem as a continuous Markov Decision Process with agent `actions' corresponding to motor movement durations, and fine tuned simultaneous independent learners for Right Ascension and Declination axes
    • Extensive indoor simulations using convolutions on laser point source for robust star detection heuristics

  • Electronic Calculus Tutor for Undergraduate Math                                                                                                 
    Advisors: Prof. Larissa Fradkin, Victor Zernov, South Bank University, London
    • Decision-tree based differentiation tutor in JAVA for differential calculus: aimed to aid teaching for dyslexics, and students with learning disabilities
    • Developed a non-deterministic question generator, enabling a larger range of practice questions; while working on previously existing MVC design-pattern
    • Applet version hosted on local website for first year undergraduate math class at university, and in consideration for proprietary release by name of 'EPACT'

  • A GO based clustering for prognosis and inference from microarray data. (pdf) (poster) (abstract) (extended abstract)
    • Extending the statistical approach using a knowledge infusion from the Gene Ontology, we propose a method that expresses a biological snapshot in terms of gene functions, and directly yields some biological understanding
    • Prognostic predictions for two B-cell lymphomas corroborate with biological literature
    • Poster presented at Asia Pacific Bioinformatics Conference (APBC), 2010
    • A Content-based Similarity Search for Monophonic Melodies. (pdf) (poster) (abstract)
      • Searching for the most similar melodies for an input query (fom a MIDI keyboard) from a musical database. We use clustering on global features followed by use of individual feature matching using Transportation distances
      • Poster presented at IBM I-CARE Workshop, 2009 held in New Delhi
    • Enhancing click-rates of online Advertisements. (pdf)
      • Local clustering for rare class analysis and SVM for classification to identify patterns that would lead to higher click rates of online advertisements
      • Adjudged best solution to problem proposed by Komli (www.komli.com)
    • Disambiguating Hindi Named Entities. (pdf) (abstract)
      • Special emphasis on the role of associated Parsargs and Verb Classes 
      • Multiple sources of evidence, word co-occurrence and the cost sensitive learning paradigm are used to get high recall and F measure values
    • Credit Card Fraud Detection: UCSD Data Mining Contest, June - July 2009
      • Detected 85.7% of fraudulent transactions (lift of 4.286 at 20% ) using J48 Decision trees and extensive feature preprocessing
      • Best Indian team (UG and PG) at UCSD 2009 Data Mining Contest
    • Amateur attempts on the Physiological Data Modeling Problem.
      • A comparitive analysis of classification algorithms on the Physiological Modeling Dataset: emphasis on the performance of Support Vector Machines and use of a semi-supervised approach in the final algorithm
    • A Color-based Indexing for Image Similarity Searching. (pdf) (abstract)

    Talks:
    • Semantically enriched tree kernels for NLP, Student Research Symposium, Language Technologies Institute, CMU. August 2013
    • Invited talk and practical workshop on `Introduction to Machine Learning', Geetanjali Institute of Technology and Science, Udaipur, 2010
    • Departmental seminar on `graph clustering' in SIGML group, IIT Kanpur, Sep  25, 2009

    Subpages (2): Abstracts Attachments
    Comments