Abhinav Mukund Kulkarni
E-mail: abhinav.kulkarni insert "at" sign uci.edu
About Me
I am a Masters student in school of Information and Computer Science at University of California, Irvine.
My research interests are in the areas bordering on both Computer
Science and Applied Statistics such as Machine Learning and Text Mining.
I completed my undergraduate education from National Institute of Technology, Tiruchirapally, India in 2009.
I worked in the Online Services Devision of Microsoft Inc., Hyderabad, India
for a little less than two years before deciding to leave for further
education. Please refer to my resume for my work at Microsoft.
Coursework
(Some of the links require valid university login)
Fall 2011: Intermediate Statistics A,
Machine Learning,
Data Structures
Winter 2012: Intermediate Statistics B,
Statistical Methodology and Data Analysis II,
Probabilistic Learning,
Information Retrieval
Spring 2012: Statistical Computing,
Graphical Models,
Computer Architecture
Fall 2012: Principles of Data Management, Statistical Methodology and Data Analysis I
Winter 2013: Bayesian Statistics
Master's Thesis
I am currently being advised by Prof. Alex Ihler for a Master's thesis. We are working on designing efficient machine learning models to evaluate Go boards and predict final territory by using local patterns and couplings between adjacent stones in the Go board.
Background: According to Wikipedia entry for Go, the game originated in China around 2,500 years back. The end goal of the game is to capture maximum possible territory. Despite its simple rules, Go has thwarted many attempts to automate it because of computational complexity and as a result traditional techniques such as alpha-beta pruning cannot be employed. Some work has been done to construct machine learning models that learn coupling between adjacent stones and make use of neighbouring terriory of a stone position.
Research Work
Complexity Evolution of Software Projects: In the summer and Fall of 2012 I worked with Prof. Crista Lopes
on the problem of analyzing open-source Java projects of varying types
and sizes to understand how they evolved over time (i.e. over multiple
releases). Currently we identify software concerns and aspects as
latent topics using LDA
and draw different matrices regarding project complexity using entropies of distributions of topic
and source code files. We were able to identify a general trend in the development of the projects by employing a simple linear regression model. Standard statistical techniques can be used to identify outlying software projects so they can be closely
studied to identify what went different in the management of those projects.
This study is continuation of an earlier work, details about which can
be found in this paper.
More details about this project and related source code can be found on GitHub.
I, part of a team of three, worked on a Kaggle
project to identify patients who would be admitted to hospitals given
medical history of previous years. We implemented various
classification and prediction algorithms to get an estimate. The
compition is still on and carries a prize pool of $3 million.
PNAS Chemistry Corpus Analysis
We did analysis of PNAS
Chemistry corpus from over two decades. Our hypothesis was that research
in Chemistry, unlike other physical sciences such as Physics and
Mathematics, is conducted in much smaller groups and scope
of collaboration is limited, in most cases, to other people in same
affiilated institutes. We found evidence that suggests our hypothesis.
Resume
PDF
Follow Me
You can follow me on LinkedIn, Facebook, Twitter and Quora.
Information and Computer Science
University of California, Irvine
Irvine, CA 92697-3425
Last modified: 15 Apr 2012
