Muthukumaran Chandrasekaran| PhD Candidate

Personal information

Date of Birth September 28, 1985
Address THINC Lab
539G Boyd GSRC
University of Georgia
Athens, GA, United States - 30602
E-mail .(JavaScript must be enabled to view this email address)
.(JavaScript must be enabled to view this email address)
Phone 706.247.2873
Degrees Held

MS Artificial Intelligence
B.Tech Mechatronics

Website Portfolio Page
Currently Working Research Assistant
THINC Lab

PhD Candidate at University of Georgia

I am Ph.D. candidate in the Computer Science department at UGA. I am a research assistant at the THINC lab working under the direction of Dr. Prashant Doshi. I received my Master's in Artificial Intelligence from UGA in 2010. Prior to that, I completed my Bachelor's in Mechatronics Engineering from SASTRA University, India in 2007.

My current research interests include multiagent planning and decision making, open agent systems, ad hoc teamwork, graphical models, and game theory. I've been serving as a peer-reviewer for top international academic conferences including AAMAS, IJCAI, and AAAI since 2009.

I love teaching and am passionate about interdisciplinary research and entrepreneurship.

Google Scholar Profile

Research Interests

  • Artificial Intelligence: Game theory, Robotics, Machine Learning,
  • Decision Making/Planning, Multi-, Open, and Ad Hoc Agent Teams
  • Exact and Approximate algorithms: I-POMDPs, Graphical (I-DIDs)
  • Evolutionary Algorithms, Natural Language Processing

Web Design Interests

  • General Web/Graphic Design and Development
  • Content Management Systems: ExpressionEngine, Wordpress
  • Databases: MySQL, MS SQLServer
  • Other: Mobile Applications(Android/iOS), Responsive Design

Resume


  • May 2008 - May 2014

    AI Researcher @THINC Lab, UGA.

    Investigate sequential planning and decision making for multi-agent settings in partially observable environments in the context of Interactive Partially Observable Markov Decision Processes (I-POMDPs) and their graphical counterparts - Interactive Dynamic Influence Diagrams (I-DIDs). Develop exact and approximate solutions and explore ways for scaling up of current methods to handle larger, more complex individual decision making problems also involving more than 2 agents. Currently investigating teamwork in finitely nested hierarchical frameworks and attempt to induce true team behavior using reinforcement learning.
  • May 2013 - Aug 2013

    Application Developer @EITS, UGA

    Designated to work mainly on projects that involve web programming. Responsibilities include 1) Program from specifications provided; using Java, Servlets, JSP and JavaScript. 2) Test against DB2 and other database structures; investigating results and generate reports. 3) Implement programs and supporting procedures efficiently. 4) Maintain appropriate documentation on assigned systems and programs.
  • Jan 2010 - May 2013

    Lead Web Developer @College of Veterinary Medicine, UGA

    Responsible for complete restructuring and migration of the College of Veterinary Medicine website (including the Teaching hospital, Academic affairs, Diagnostic lab and all academic departments) to a content management system called ExpressionEngine2.0. Also responsible for maintaining, updating existing websites and inspecting them for errors or outdated content and perform quality control reviews of new sites and participate in strategic and operational planning.
  • May 2012 - Aug 2012

    Data Management Intern/Consultant @Synovus Financial

    Designated to work on the Master Data Management project. Mission - Research the data quality issues surrounding the 'Master' data and provide a solution to integrate data from several source systems to a central repository and give the ability to Synovus' user population to maintain integral data that is not maintained on the source systems through custom user interface screens. Successfully implemented the project and transitioned the knowledge to Synovus' internal team. I was requested to provide consulting and help during the ongoing production phase even after the internship ended in August.
  • Aug 2010 - Dec 2010

    Teaching Assistant @CS Department, UGA

    Instructor for CSCI 1100: Introduction to Personal Computing Lab. Also responsible for grading, leading group discussions and holding office hours to interact with students and helping them with assignments and homework.

Skills Set

Programming Java C/C++ Matlab Prolog Lisp Python
Languages English Hindi Tamil Arabic
Web Development HTML CSS JavaScript JQuery PHP ColdFusion JSP Servlets Flash/AS3 SEO
CMS ExpressionEngine WordPress Joomla
Database MySQL SQLServer DB2 PostgreSQL Oracle MS Access
Operating System Windows 7/XP/Vista Linux Redhat/Ubuntu Mac
Tools/Platforms Hugin Expert Netica NeuroShell WEKA Photoshop Protege Poser MimicPro
Other MS MDS Sharepoint OWL/RDF Google APIs Facebook APIs AutoCAD
Certificates 3D Studio Max Maya

Relevant Coursework

AI Core Symbolic Programming Deductive Systems Artificial Intelligence
Decision Making Decision Making under Uncertainty Judgement and Decision Making
CS Core Database Management Computer Networks Algorithms
Nature Inspired Optimization Evolutionary Computation Computational Intelligence
Advanced AI Machine Learning Natural Language Processing
Advanced CS Advanced Distributed Systems AI and the Web Image Processing and Computer Graphics Knowledge-Based Systems
Other Coursework Statistical Methods Model Theory Academic Writing

Education

  • Aug 2010 - May 2014

    University of Georgia, USA

    • PhD Candidate in Computer Science
    • Major AdvisorDr. Prashant Doshi
    • Awards - Full Scholarship & Stipend through a Graduate Assistantship
    • GPA - 3.86/4.0
  • Aug 2007 - May 2010

    University of Georgia, USA

    • MS in Artificial Intelligence
    • Major AdvisorDr. Prashant Doshi
    • Awards - Full Scholarship & Stipend through a Graduate Assistantship
    • ThesisApproximate Model Equivalence for Interactive Dynamic Influence Diagram [ Article ]
    • GPA - 3.22/4.0
  • Aug 2003 - May 2007

    SASTRA University, India

    • B.Tech in Mechatronics
    • Major AdvisorDr. Aushitos Das
    • Awards - Best Outgoing Student and Dean's List (top 10%) for 4 straight years
    • ThesisRF Remote Controlled Surveillance Robot (Range: 40ft)
    • GPA - 8.90/10 → 3.70/4.0

Conference Publications

  • 2016
    Can Bounded and Self-Interested Agents be Teammates?
    • Muthukumaran Chandrasekaran, Prashant Doshi, Yifeng Zeng, Yingke Chen, Can Bounded and Self-Interested Agents be Teammates? Application to Planning in Ad Hoc Teams, Under review in Journal of Autonomous Agents and Multiagent Systems (JAAMAS), 2016.
  • 2016
    Individual Planning in Typed and Open Multiagent Systems
    • Muthukumaran Chandrasekaran, Adam Eck, Prashant Doshi, Leenkiat Soh, Individual Planning in Typed and Open Multiagent Systems, to appear in International Conference on Uncertainty in Artificial Intelligence (UAI-16), 2016. [ Article ]
  • 2016
    Bayesian Markov Games with Explicit Finite-Level Types
    • Muthukumaran Chandrasekaran, Yingke Chen, Prashant Doshi,Bayesian Markov Games with Explicit Finite-Level Types, submitted to NIPS, 2016.
  • 2016
    Bayesian Markov Games with Explicit Finite-Level Types (Student Abstract)
    • Muthukumaran Chandrasekaran, Yingke Chen, Prashant Doshi,Bayesian Markov Games with Explicit Finite-Level Types, Student Abstract, in AAAI, 2016.
  • 2016
    Bayesian Markov Games with Explicit Finite-Level Types (Workshop)
    • Muthukumaran Chandrasekaran, Yingke Chen, Prashant Doshi,Bayesian Markov Games with Explicit Finite-Level Types, AAAI Workshop on Multiagent Interaction Without Prior Coordination (MIPC), 2016.
  • 2015
    Approximating Behavioral Equivalence for Scaling Solutions of I-DIDs
    • Yifeng Zeng, Prashant Doshi, Yingke Chen, Yinghui Pan, Hua Mao, Muthukumaran Chandrasekaran, Approximating Behavioral Equivalence for Scaling Solutions of I-DIDs, In Knowledge and Information Systems (KAIS), 2015. [ Article ]
  • 2014
    Team Behavior in Interactive Dynamic Influence Diagrams with Applications to Ad Hoc Teams
    • Muthukumaran Chandrasekaran, Prashant Doshi, Yifeng Zeng, Yingke Chen, Team Behavior in Interactive Dynamic Influence Diagrams with Applications to Ad Hoc Teams, AAMAS Workshop on Multi-agent Sequential Decision Making (MSDM), 2014. [ Article ]
  • 2014
    Team Behavior in Interactive Dynamic Influence Diagrams with Applications to Ad Hoc Teams (Extended Abstract)
    • Muthukumaran Chandrasekaran, Prashant Doshi, Yifeng Zeng, Yingke Chen, Team Behavior in Interactive Dynamic Influence Diagrams with Applications to Ad Hoc Teams (Extended Abstract), in AAMAS, 2014. [ Article ]
  • 2012
    Heuristics to Extract the Main Text from a Captured Web Page
    • Muthukumaran Chandrasekaran, Michael Covington, Heuristics to Extract the Main Text from a Captured Web Page, In Proceedings of the International Conference on Artificial Intelligence, 2012. [ Article ]
  • 2011
    Evolving Efficient Sensor Arrangement and Obstacle Avoidance Control Logic for a Miniature Robot
    • Muthukumaran Chandrasekaran, Karthik Nadig, and Khaled Rasheed, Evolving Efficient Sensor Arrangement and Obstacle Avoidance Control Logic for a Miniature Robot, In Proceedings of the Twenty-Fourth International Conference on Industrial Engineering and Other Applications of Applied Intelligent Systems (IEA/AIE-11), 2011. [2nd Best Research Paper Award] [ Article ]
  • 2011
    Utilizing Partial Policies for Identifying Equivalence of Behavioral Models
    • Yifeng Zeng, Prashant Doshi, Yinghui Pan, Hua Mao, Muthukumaran Chandrasekaran, Jian Luo, Utilizing Partial Policies for Identifying Equivalence of Behavioral Models, In Proceedings of the Twenty-Fifth AAAI Conference on Artificial Intelligence (AAAI-11), pp.1083-1088, 2011. [ Article ]
  • 2010
    Approximate Solutions of Interactive Dynamic Influence Diagrams Using ε-Behavioral Equivalence
    • Muthukumaran Chandrasekaran, Prashant Doshi, Yifeng Zeng, Approximate Solutions of Interactive Dynamic Influence Diagrams Using ε-Behavioral Equivalence, In 11th International Symposium on Artificial Intelligence and Mathematics (ISAIM-10), 2010. [ Article ]
  • 2010
    ε-Subjective Equivalence of Models for Interactive Dynamic Influence Diagrams
    • Prashant Doshi, Muthukumaran Chandrasekaran, Yifeng Zeng, ε-Subjective Equivalence of Models for Interactive Dynamic Influence Diagrams, In Proceedings of 2010 IEEE/WIC/ACM International Conference on Intelligent Agent Technology (IAT-10), pp. 165-172, 2010. [ Article ]
  • 2008
    Path Normalcy Analysis using Nearest Neighbor Outlier Detection
    • David Luper, Muthukumaran Chandrasekaran, Khaled Rasheed, Hamid R. Arabnia, Path Normalcy Analysis using Nearest Neighbor Outlier Detection, In Proceedings of the International Conference on Artificial Intelligence (ICAI-08), pp. 776-783, 2008. [ Article ]

Posters/Presentations

  • 2013
    Learning Team Behavior Using Individual Decision Making in Multiagent Settings Using Interactive DIDs
    • Muthukumaran Chandrasekaran, Learning Team Behavior Using Individual Decision Making in Multiagent Settings Using Interactive DIDs, at CS Research Day Poster Competition, Computer Science Department, UGA, 2013. [ Talk ]
  • 2012
    Using Reinforcement Learning to Model True Team Behavior in Uncertain Multiagent Settings in Interactive DIDs
    • Muthukumaran Chandrasekaran, Using Reinforcement Learning to Model True Team Behavior in Uncertain Multiagent Settings in Interactive DIDs, at CS Research Day Poster Competition, Computer Science Department, UGA, 2012. [ Talk ]
  • 2011
    Solving Multi-Agent Sequential Decision Making Problems Using Interactive Dynamic Influence Diagrams
    • Muthukumaran Chandrasekaran, Solving Multi-Agent Sequential Decision Making Problems Using Interactive Dynamic Influence Diagrams, at CS Research Day Poster Competition Computer Science Department, UGA, 2011. [3rd Best Poster Award & $250 Cash] [ Talk ]
  • 2008
    Information-Theoretic Approach to Model Identification in Interactive Influence Diagrams
    • Yifeng Zeng and Prashant Doshi, presented by Muthukumaran Chandrasekaran, Information-Theoretic Approach to Model Identification in Interactive Influence Diagrams, In Proceedings of IEEE WIC/ACM IAT, Australia, 2008. [ Article ] [ Talk ]

References

Prashant Doshi

Major Advisor (5+ Years)
  • Associate Professor
  • THINC Lab, CS Dept, UGA
  • Email: .(JavaScript must be enabled to view this email address)

Yifeng Zeng

Major Advisor (5+ Years)
  • Reader
  • Teesside University, UK
  • Email: .(JavaScript must be enabled to view this email address)

Kat Gilmore

Supervisor (2 Years)
  • Director of Public Relations
  • College of Vet Med, UGA
  • Email: .(JavaScript must be enabled to view this email address)

Shoba Rao

Supervisor (3 Months)
  • Sr. Director of Data Mgmt
  • Synovus Bank, Columbus-GA
  • Email: .(JavaScript must be enabled to view this email address)

Jason Davis

Freelance Client (4 Years)
  • Founder/CEO
  • Picobarn LLC.
  • Email: .(JavaScript must be enabled to view this email address)

J. Lane Chasteen

Supervisor (3 Months)
  • Application Analyst Specialist
  • EITS, UGA
  • Email: .(JavaScript must be enabled to view this email address)

Wayne Willmon

Partner (6 Months)
  • Database Administrator
  • EITS, UGA
  • Email: .(JavaScript must be enabled to view this email address)

Testimonials

"It is a pleasure to work with Muthu! He takes direction well, works well independently, displays initiative, offers his own great feedback and ideas on collaborative projects, and is a true team player! Anyone who would like more detail may contact me at kygilmor@uga.edu. Bottom line: If I could keep Muthu forever, I would!"August 30, 2010

Kat Gilmore
PR Director @ CVM, UGA

"Muthu was hired as a summer intern to work on the Master Data Management project at Synovus. Muthu did a thorough research, came up with solutions, installed, configured the MS MDS product and developed many user interface screens. Muthu worked day and night to successfully implement the project and transition the knowledge to Synovus internal team. I would hire Muthu again without hesitation."August 12, 2012

Shoba Rao
Sr. Director @ Synovus

"Muthu is a driven an energetic problem-solver. He approaches problems with an open mind and does not jump to conclusions. His method is logical, and his ability to foresee potential problems is only exceeded by his ability to elegantly solve them."February 14, 2011

Jason Davis
Founder/CEO @ Picobarn LLC.

"We were privileged to hire Muthu when our part-time developer left for a summer internship. He is dedicated, works hard, and is self motivated. He had to quickly learn our systems and processes. He exceeded our expectations and was able to contribute, significantly, to several of our projects."December 17, 2013

J. Lane Chasteen
Appl Analyst Spec @ EITS, UGA

Projects


  • Jan 2015

    Project Domain:

    Big Data, Mining Massive Datasets, Artificial Intelligence, Machine Learning

    Project Description:

    We introduce a novel approach to mining and leveraging data concerning stocks in order to predict the performance of new stocks following their initial public offering, a traditionally difficult task due to the lack of information and historical performance data. We collect a large corpus of articles for every existing stock between March 1st, 2014 and March 1st, 2015. We create weighted feature vectors by calculating the TF-IDF values for every word that appears in a document about a given stock. We then perform locality-sensitive hashing on these feature vectors and bucket similar stocks into a neighborhood. Using a variety of monotonically decreasing functions, we examine the relationship between 30-day price fluctuations of each IPOs neighbor by propagating each neighbor’s 30-day change to the IPO relative to its distance. Locality-sensitive hashing uncovers interesting relationships between IPOs and nearby neighbors. We observe a strong correlation between neighbor stock distance and respective 30-day performance. .. Learn more

  • May 2014

    Project Domain:

    Crowd Sourcing, Handwriting OCR, Software Development

    Project Description:

    The goal of this project is to develop a note-taking software which will take as input, scanned handwritten notes shared from a “smart board” installed in educational institutions (and the like) including schools, universities, conference rooms etc. The software will potentially allow students or colleagues to search, categorize, and tag these notes shared in real-time in order to allow for easy navigation. For this, the software needs to be able to understand the content of these notes. The first task will be to identify/recognize the handwriting of a teacher/presenter who used the smart board to share his/her notes. Obviously, utility of such software will be contingent on the existence of a compatible “smart board” with real-time note-sharing abilities. .. Learn more

  • Feb 2014

    Project Domain:

    WordPress, UMD Framework (faculty/staff directory),
    PHP, HTML, JS/JQuery, CSS, Dreamweaver

    Project Description:

    Vaccines and Biological Countermeasures Interest Group Website: http://vaccines.uga.edu/
    Pathogens and Immunity Interest Group Website: http://vaccines.uga.edu/

    These sites went live on 2/7/2014. These sites are powered by a content management system called WordPress.

  • Aug 2013

    Project Domain:

    WordPress, PHP, MySQL

    Project Description:

    Website: http://daysinnathens.com/

    This site went live on 8/11/2013. This site is powered by a content management system called WordPress.

  • Aug 2013

    Project Domain:

    HTML, CSS, JavaScript/JQuery, Photoshop, Dreamweaver,
    Longwell (RDF-Based browser for Publications)

    Project Description:

    Website: http://thinc.cs.uga.edu/

    This site went live on 8/11/2013. This was a voluntary undertaking and a pleasant surprise to the THINC lab director and my major advisor, Dr Prashant Doshi.

  • Mar 2013

    Project Domain:

    WordPress, HTML, CSS, JQuery, Javascript, PHP, MySQL, Facebook API

    Project Description:

    Website: http://www.phil-athens.com/

  • Mar 2013

    Project Domain:

    ExpressionEngine, HTML, CSS, JQuery, Javascript, PHP

    Project Description:

    Website: http://vet.uga.edu/

  • Mar 2013

    Project Domain:

    Responsive Web Design, ExpressionEngine, HTML, CSS, JS/JQuery

    Project Description:

    Website: http://muthuchandrasekaran.com/

    This started as a responsive web design experiment and I loved the way it looked, so I turned this into my web portfolio which I keep up to date using a content management system called ExpressionEngine.

  • Dec 2012

    Project Domain:

    ExpressionEngine, HTML, CSS, Javascript, JQuery

    Project Description:

    Website: http://dime4dollar.com/

  • Mar 2012

    Project Domain:

    AI and the Web, Natural Language Processing, Text Parsing, Web Services, Ontologies

    Project Description:

    Many companies use existing web based systems like Taleo as their job application interface. Some systems allow the applicant to upload their resume and parse it to automatically populate the fields in the application. However, these systems do a poor job in populating the fields accurately and sometimes require extensive user interaction. They often store the information in flat files or databases only. Our objective was to create an Automated Human Resources Tool, that would take in a text base resume as input, parse the contents, render it machine readable and efficiently populate data into corresponding fields. For this we created effective parsing rules for automating the data collection, and used ontologies for knowledge representation. We exposed our application through via RESTful Web Services for platform independence. .. Learn more

  • Dec 2011

    Project Domain:

    Distributed Robotics, Artificial Intelligence, Team work, Multi-Agent Systems

    Project Description:

    We will simulate an environment where a robot team is deployed in a train wreck situation and each robot is faced with the task of individually exploring areas of a crash site and reporting as much information as possible about the site to a remote master robot which will then regroup the team and attempt to reconstruct the map from the individually acquired partial maps which over time, will result in an accurate reconstruction of the entire map and reassign jobs to them based on the current information available to it. Assuming there is a cost for communication, we will study the trade-off between the efficiency of the map merging algorithm and the speed with which it is carried out. We will optimize the number of robots needed to efficiently carryout the map-merging algorithm assuming there is a cost for including each robot in the team. We will also compare the performance of our algorithm with existing map-merging algorithms in terms of the costs defined above, time taken, accuracy of the reconstructed map, and the efficiency of the algorithm. .. Learn more

  • Aug 2011

    Project Domain:

    HTML, CSS, PHP, JavaScript/JQuery

    Project Description:

    Website: http://isa.uga.edu/

    I designed and developed the entire website for the Indian Students Association (ISA) as a newly recruited Web Master for the ISA Committee of 2011 – 2012. This was a voluntary project and the website went live on Aug 15, 2011.

  • Jun 2011

    Project Domain:

    HTML, Flash AS3, CSS, PHP, JavaScript/JQuery, Dreamweaver, Photoshop

    Project Description:

    Website: http://www.nit-mut.com/

    This website is now offline. I built and designed this site and all its graphics from scratch as a surprise to my wife for our wedding. This website included functionality like guest book, RSVP, music etc.

  • May 2011

    Project Domain:

    Distributed Artificial Intelligence, Multi-Agent Systems, Decision making under Uncertainty, Agent Modeling, Probabilistic Reasoning, Behavioral Equivalences, Graphical Models: I-DIDs

    Project Description:

    We present a novel approach for identifying exact and approximate behavioral equivalence between models of agents. This is significant because both decision making and game play in multi-agent settings must contend with behavioral models of other agents in order to predict their actions. One approach that reduces the complexity of the model space is to group models that are behaviorally equivalent. Identifying equivalence between models requires solving them and comparing entire policy trees. Because the trees grow exponentially with the horizon, our approach is to focus on partial policy trees for comparison and determining the distance between updated beliefs at the leaves of the trees. We propose a principled way to determine how much of the policy trees to consider, which trades off solution quality for efficiency. We investigate this approach in the context of the interactive dynamic influence diagram and evaluate its performance. .. Learn more

  • May 2011

    Project Domain:

    RESTful Web Services, POMDP, Agent Technology, Decision Making under Uncertainty

    Project Description:

    Several POMDP resolution toolkits are available from various locations: Cassandra's POMDP Toolkit (for exact solutions) and Approximate POMDP Planning Toolkit (APPL, for approximate solutions). Robust applications streamlining and centralizing these functions and programs are largely not available. Our project offers a central location to solve POMDPs using a number of resolution techniques. As the first centralized, web-based POMDP solver, our project offers several improvements over pre-existing solutions. Our solver is (1) Platform Independent: Utilizing basic JavaScript/AJAX and HTML on the client side, our solution can be run on virtually any PC, (2) Scalable: By using web services, we can point service calls to any number of machines, which can be added in the future, and (3) Easy to Use: Providing web services to dedicated machines pre-configured with toolkits, the user need only to provide POMDP files and some parameters to solve their POMDPs. .. Learn more

  • Feb 2011

    Project Domain:

    ColdFusion, MS SQLServer, Flash AS3, HTML, CSS, Javascript, JQuery

    Project Description:

    Website: http://www.40gallonchallenge.org/

  • Dec 2010

    Project Domain:

    Database Management Systems, MySQL, Struts 2, UML Design

    Project Description:

    We created a new and improved student registration system for the University of Georgia. We called it NEWASIS. The idea for the project came out of necessity: the current registration system, OASIS, has a whole range of problems it faces, from the database backend which shuts down daily at 2:00 AM for optimization purposes, to the graphical user interface, which is clumsy and unintuitive. The features of this new system include a better looking UI with improved search capability and a more efficiently designed database structure with ACID properties for transactions and other general properties such as high availability and high reliability. It was built on a java-based Struts 2.0 web framework. This project also involved creating UML designs for the database.

  • Dec 2010

    Project Domain:

    Robotics, Genetic Algorithms, Genetic Programming

    Project Description:

    We evaluated the use of Genetic Programming (GP) to evolve a controller that implements an Obstacle Avoidance (OA) behavior in a miniature robot. The GP system generates the OA logic equation offline on a simulated dynamic 2-D environment that transforms the sensory inputs from a simulated robot to a controller decision. The set of motor control decisions for all possible sensor trigger sequences is applied to a real robot which is then tested on a real environment. Needless to say, the efficiency of this OA robot depends on the information it can receive from its surroundings. This information is dependent on the sensor module design. Thus, we also present a Genetic Algorithm (GA) that evolves a sensor arrangement taking into consideration economic issues as well as the usefulness of the information that can be retrieved. The performance of our algorithm is compared with that of a hybrid neural network and also with an online (real time) evolution method. .. Learn more

  • May 2010

    Project Domain:

    Distributed Artificial Intelligence, Multi-Agent Systems, Decision making under Uncertainty, Agent Modeling, Probabilistic Reasoning, Behavioral Equivalences, Graphical Models: I-DIDs

    Project Description:

    Interactive dynamic influence diagrams (I-DIDs) graphically visualize a sequential decision problem for uncertain settings where multiple agents interact not only amongst themselves but also with the environment that they are in. Algorithms currently available for solving these I-DIDs face the issue of an exponentially growing candidate model space ascribed to the other agents, over time. One such algorithm identifies and prunes behaviorally equivalent models and replaces them with a representative thereby reducing the model space. We seek to further reduce the complexity by additionally pruning models that are approximately subjectively equivalent. Toward this, we define subjective equivalence in terms of the distribution over the subject agent's future action observation paths, and introduce the notion of o;-subjective equivalence. We present a new approximation technique that uses our new definition of subjective equivalence to reduce the candidate model space by pruning models that are o;-subjectively equivalent with representative ones. .. Learn more

  • Aug 2009

    Project Domain:

    Perl, PHP, MySQL, Google Adwords API, Custom Search API, Dreamweaver

    Project Description:

    Website: http://picobarn.com/

    Consulted for Picobarn. Built a Keyword Analysis Tool and the Picobarn Index - an improvement over the KEI index to measure goodness of keywords based on their search volume and competition.

  • May 2009

    Project Domain:

    Natural Language Processing, Machine Learning

    Project Description:

    Web pages generally contain a large amount of information that is not part of their main content, such as navigation panels, copyright and privacy notices, and advertisements. Such irrelevant information, known as Webpage noise, can seriously affect the accuracy of search and other web mining applications; the act of removing it is called "cleaning." In this paper, we introduce a different approach based on heuristics that extract the main text on the basis of its characteristics rather than any previously recognized structural pattern. The key insight is that the main text consists of blocks of relatively long lines that include high-frequency (common) words, while headlines and "noise" consist of smaller amounts of text, isolated from each other, with a higher proportion of rare words. .. Learn more

  • Dec 2008

    Project Domain:

    Problem Solving, Propositional Logic, Logic Programming, PROLOG Programming

    Project Description:

    Kakuro puzzles can be solved using a brute force strategy by converting the puzzle into equations with variables representing each unfilled square. However, a brute force methodology will not work efficiently for large puzzles. It is also observed that it would not just suffice to solve the equations in the order in which they originally appeared. I implemented four major algorithms that carry out a preprocessing step to re-order the equations before actually solving them. These algorithms were compared to the brute force method and drastic improvements in terms of time taken to solve each of 172 different puzzles are shown. Also, created operators in Prolog corresponding to the logical connectives of propositional logic and implemented several predicates for manipulating well-formed formulas and verifying their properties.

  • May 2008

    Project Domain:

    Artificial Neural Networks, Probabilistic Neural Networks, Classification Problem

    Project Description:

    This study compared two alternative artificial neural network architectures for predicting forest cover types. The experiments evaluated four wilderness areas in the Roosevelt National Forest, located in the Front Range of northern Colorado. The data set was acquired from US Forest Service inventory information and from the standard digital spatial data processed in a geographic information system. The results of the comparison indicated that a probabilistic neural network model predicted the forest cover type more accurately than a three-layer feed forward back propagation network model.

  • May 2008

    Project Domain:

    Genetic Algorithms, Hill Climbing, Particle Swarm & Raindrop Optimization, Simulated Annealing

    Project Description:

    I implemented Hill Climbing and Raindrop Optimization algorithms to solve a scheduling problem in the forestry domain. I also implemented Genetic Algorithms, Simulated Annealing, and Particle Swarm Optimization to solve benchmark problems such as Multiple Fault Diagnosis, Mobile Subscriber Equipment, Schaffer's function, and the Snake in the Box.

  • May 2008

    Project Domain:

    Machine Learning, Data Mining

    Project Description:

    We presented a machine learning technique that recognizes patterns of movement, using GPS data and time stamps, to gain the ability to detect regions of time containing abnormal movement. We used features extracted from this raw GPS data to describe a person's movement over discrete regions of time. Then, we used a nearest neighbor approach to determine outliers in a distribution of time regions. We produced a distance range value for a distribution in conjunction with normalized scores depicting the degree to which each time region movement consistent with other time regions being analyzed. We also produced a classification of each day as normal or abnormal. .. Learn more

  • May 2007

    Project Domain:

    Robotics, Electronics

    Project Description:

    I built a remote controlled robotic car capable of audio and video surveillance, with a transmission range of up to 40 feet. The brick-sized robot runs on a 9V battery pack. The wheels are actuated by 2 DC motors and its maneuvering is controlled using a remote that transmits control signals to the robot through RF waves of a chosen frequency. The robot houses a receiver for receiving the control signals from the remote and a transmitter for sending audio/video signals from the wireless camera (with a built-in microphone) fitted on the robot chassis to an output device such as a TV monitor tuned to the frequency of transmission.

PhD Projects

  • Jan 2015

    Project Domain:

    Big Data, Mining Massive Datasets, Artificial Intelligence, Machine Learning

    Project Description:

    We introduce a novel approach to mining and leveraging data concerning stocks in order to predict the performance of new stocks following their initial public offering, a traditionally difficult task due to the lack of information and historical performance data. We collect a large corpus of articles for every existing stock between March 1st, 2014 and March 1st, 2015. We create weighted feature vectors by calculating the TF-IDF values for every word that appears in a document about a given stock. We then perform locality-sensitive hashing on these feature vectors and bucket similar stocks into a neighborhood. Using a variety of monotonically decreasing functions, we examine the relationship between 30-day price fluctuations of each IPOs neighbor by propagating each neighbor’s 30-day change to the IPO relative to its distance. Locality-sensitive hashing uncovers interesting relationships between IPOs and nearby neighbors. We observe a strong correlation between neighbor stock distance and respective 30-day performance. .. Learn more

  • May 2014

    Project Domain:

    Crowd Sourcing, Handwriting OCR, Software Development

    Project Description:

    The goal of this project is to develop a note-taking software which will take as input, scanned handwritten notes shared from a “smart board” installed in educational institutions (and the like) including schools, universities, conference rooms etc. The software will potentially allow students or colleagues to search, categorize, and tag these notes shared in real-time in order to allow for easy navigation. For this, the software needs to be able to understand the content of these notes. The first task will be to identify/recognize the handwriting of a teacher/presenter who used the smart board to share his/her notes. Obviously, utility of such software will be contingent on the existence of a compatible “smart board” with real-time note-sharing abilities. .. Learn more

  • Mar 2012

    Project Domain:

    AI and the Web, Natural Language Processing, Text Parsing, Web Services, Ontologies

    Project Description:

    Many companies use existing web based systems like Taleo as their job application interface. Some systems allow the applicant to upload their resume and parse it to automatically populate the fields in the application. However, these systems do a poor job in populating the fields accurately and sometimes require extensive user interaction. They often store the information in flat files or databases only. Our objective was to create an Automated Human Resources Tool, that would take in a text base resume as input, parse the contents, render it machine readable and efficiently populate data into corresponding fields. For this we created effective parsing rules for automating the data collection, and used ontologies for knowledge representation. We exposed our application through via RESTful Web Services for platform independence. .. Learn more

  • Dec 2011

    Project Domain:

    Distributed Robotics, Artificial Intelligence, Team work, Multi-Agent Systems

    Project Description:

    We will simulate an environment where a robot team is deployed in a train wreck situation and each robot is faced with the task of individually exploring areas of a crash site and reporting as much information as possible about the site to a remote master robot which will then regroup the team and attempt to reconstruct the map from the individually acquired partial maps which over time, will result in an accurate reconstruction of the entire map and reassign jobs to them based on the current information available to it. Assuming there is a cost for communication, we will study the trade-off between the efficiency of the map merging algorithm and the speed with which it is carried out. We will optimize the number of robots needed to efficiently carryout the map-merging algorithm assuming there is a cost for including each robot in the team. We will also compare the performance of our algorithm with existing map-merging algorithms in terms of the costs defined above, time taken, accuracy of the reconstructed map, and the efficiency of the algorithm. .. Learn more

  • May 2011

    Project Domain:

    Distributed Artificial Intelligence, Multi-Agent Systems, Decision making under Uncertainty, Agent Modeling, Probabilistic Reasoning, Behavioral Equivalences, Graphical Models: I-DIDs

    Project Description:

    We present a novel approach for identifying exact and approximate behavioral equivalence between models of agents. This is significant because both decision making and game play in multi-agent settings must contend with behavioral models of other agents in order to predict their actions. One approach that reduces the complexity of the model space is to group models that are behaviorally equivalent. Identifying equivalence between models requires solving them and comparing entire policy trees. Because the trees grow exponentially with the horizon, our approach is to focus on partial policy trees for comparison and determining the distance between updated beliefs at the leaves of the trees. We propose a principled way to determine how much of the policy trees to consider, which trades off solution quality for efficiency. We investigate this approach in the context of the interactive dynamic influence diagram and evaluate its performance. .. Learn more

  • May 2011

    Project Domain:

    RESTful Web Services, POMDP, Agent Technology, Decision Making under Uncertainty

    Project Description:

    Several POMDP resolution toolkits are available from various locations: Cassandra's POMDP Toolkit (for exact solutions) and Approximate POMDP Planning Toolkit (APPL, for approximate solutions). Robust applications streamlining and centralizing these functions and programs are largely not available. Our project offers a central location to solve POMDPs using a number of resolution techniques. As the first centralized, web-based POMDP solver, our project offers several improvements over pre-existing solutions. Our solver is (1) Platform Independent: Utilizing basic JavaScript/AJAX and HTML on the client side, our solution can be run on virtually any PC, (2) Scalable: By using web services, we can point service calls to any number of machines, which can be added in the future, and (3) Easy to Use: Providing web services to dedicated machines pre-configured with toolkits, the user need only to provide POMDP files and some parameters to solve their POMDPs. .. Learn more

  • Dec 2010

    Project Domain:

    Database Management Systems, MySQL, Struts 2, UML Design

    Project Description:

    We created a new and improved student registration system for the University of Georgia. We called it NEWASIS. The idea for the project came out of necessity: the current registration system, OASIS, has a whole range of problems it faces, from the database backend which shuts down daily at 2:00 AM for optimization purposes, to the graphical user interface, which is clumsy and unintuitive. The features of this new system include a better looking UI with improved search capability and a more efficiently designed database structure with ACID properties for transactions and other general properties such as high availability and high reliability. It was built on a java-based Struts 2.0 web framework. This project also involved creating UML designs for the database.

  • Dec 2010

    Project Domain:

    Robotics, Genetic Algorithms, Genetic Programming

    Project Description:

    We evaluated the use of Genetic Programming (GP) to evolve a controller that implements an Obstacle Avoidance (OA) behavior in a miniature robot. The GP system generates the OA logic equation offline on a simulated dynamic 2-D environment that transforms the sensory inputs from a simulated robot to a controller decision. The set of motor control decisions for all possible sensor trigger sequences is applied to a real robot which is then tested on a real environment. Needless to say, the efficiency of this OA robot depends on the information it can receive from its surroundings. This information is dependent on the sensor module design. Thus, we also present a Genetic Algorithm (GA) that evolves a sensor arrangement taking into consideration economic issues as well as the usefulness of the information that can be retrieved. The performance of our algorithm is compared with that of a hybrid neural network and also with an online (real time) evolution method. .. Learn more

MS Projects

  • May 2010

    Project Domain:

    Distributed Artificial Intelligence, Multi-Agent Systems, Decision making under Uncertainty, Agent Modeling, Probabilistic Reasoning, Behavioral Equivalences, Graphical Models: I-DIDs

    Project Description:

    Interactive dynamic influence diagrams (I-DIDs) graphically visualize a sequential decision problem for uncertain settings where multiple agents interact not only amongst themselves but also with the environment that they are in. Algorithms currently available for solving these I-DIDs face the issue of an exponentially growing candidate model space ascribed to the other agents, over time. One such algorithm identifies and prunes behaviorally equivalent models and replaces them with a representative thereby reducing the model space. We seek to further reduce the complexity by additionally pruning models that are approximately subjectively equivalent. Toward this, we define subjective equivalence in terms of the distribution over the subject agent's future action observation paths, and introduce the notion of o;-subjective equivalence. We present a new approximation technique that uses our new definition of subjective equivalence to reduce the candidate model space by pruning models that are o;-subjectively equivalent with representative ones. .. Learn more

  • May 2009

    Project Domain:

    Natural Language Processing, Machine Learning

    Project Description:

    Web pages generally contain a large amount of information that is not part of their main content, such as navigation panels, copyright and privacy notices, and advertisements. Such irrelevant information, known as Webpage noise, can seriously affect the accuracy of search and other web mining applications; the act of removing it is called "cleaning." In this paper, we introduce a different approach based on heuristics that extract the main text on the basis of its characteristics rather than any previously recognized structural pattern. The key insight is that the main text consists of blocks of relatively long lines that include high-frequency (common) words, while headlines and "noise" consist of smaller amounts of text, isolated from each other, with a higher proportion of rare words. .. Learn more

  • Dec 2008

    Project Domain:

    Problem Solving, Propositional Logic, Logic Programming, PROLOG Programming

    Project Description:

    Kakuro puzzles can be solved using a brute force strategy by converting the puzzle into equations with variables representing each unfilled square. However, a brute force methodology will not work efficiently for large puzzles. It is also observed that it would not just suffice to solve the equations in the order in which they originally appeared. I implemented four major algorithms that carry out a preprocessing step to re-order the equations before actually solving them. These algorithms were compared to the brute force method and drastic improvements in terms of time taken to solve each of 172 different puzzles are shown. Also, created operators in Prolog corresponding to the logical connectives of propositional logic and implemented several predicates for manipulating well-formed formulas and verifying their properties.

  • May 2008

    Project Domain:

    Artificial Neural Networks, Probabilistic Neural Networks, Classification Problem

    Project Description:

    This study compared two alternative artificial neural network architectures for predicting forest cover types. The experiments evaluated four wilderness areas in the Roosevelt National Forest, located in the Front Range of northern Colorado. The data set was acquired from US Forest Service inventory information and from the standard digital spatial data processed in a geographic information system. The results of the comparison indicated that a probabilistic neural network model predicted the forest cover type more accurately than a three-layer feed forward back propagation network model.

  • May 2008

    Project Domain:

    Genetic Algorithms, Hill Climbing, Particle Swarm & Raindrop Optimization, Simulated Annealing

    Project Description:

    I implemented Hill Climbing and Raindrop Optimization algorithms to solve a scheduling problem in the forestry domain. I also implemented Genetic Algorithms, Simulated Annealing, and Particle Swarm Optimization to solve benchmark problems such as Multiple Fault Diagnosis, Mobile Subscriber Equipment, Schaffer's function, and the Snake in the Box.

  • May 2008

    Project Domain:

    Machine Learning, Data Mining

    Project Description:

    We presented a machine learning technique that recognizes patterns of movement, using GPS data and time stamps, to gain the ability to detect regions of time containing abnormal movement. We used features extracted from this raw GPS data to describe a person's movement over discrete regions of time. Then, we used a nearest neighbor approach to determine outliers in a distribution of time regions. We produced a distance range value for a distribution in conjunction with normalized scores depicting the degree to which each time region movement consistent with other time regions being analyzed. We also produced a classification of each day as normal or abnormal. .. Learn more

B.Tech Projects

  • May 2007

    Project Domain:

    Robotics, Electronics

    Project Description:

    I built a remote controlled robotic car capable of audio and video surveillance, with a transmission range of up to 40 feet. The brick-sized robot runs on a 9V battery pack. The wheels are actuated by 2 DC motors and its maneuvering is controlled using a remote that transmits control signals to the robot through RF waves of a chosen frequency. The robot houses a receiver for receiving the control signals from the remote and a transmitter for sending audio/video signals from the wireless camera (with a built-in microphone) fitted on the robot chassis to an output device such as a TV monitor tuned to the frequency of transmission.

Web-Based Projects

  • Feb 2014

    Project Domain:

    WordPress, UMD Framework (faculty/staff directory),
    PHP, HTML, JS/JQuery, CSS, Dreamweaver

    Project Description:

    Vaccines and Biological Countermeasures Interest Group Website: http://vaccines.uga.edu/
    Pathogens and Immunity Interest Group Website: http://vaccines.uga.edu/

    These sites went live on 2/7/2014. These sites are powered by a content management system called WordPress.

  • Aug 2013

    Project Domain:

    WordPress, PHP, MySQL

    Project Description:

    Website: http://daysinnathens.com/

    This site went live on 8/11/2013. This site is powered by a content management system called WordPress.

  • Aug 2013

    Project Domain:

    HTML, CSS, JavaScript/JQuery, Photoshop, Dreamweaver,
    Longwell (RDF-Based browser for Publications)

    Project Description:

    Website: http://thinc.cs.uga.edu/

    This site went live on 8/11/2013. This was a voluntary undertaking and a pleasant surprise to the THINC lab director and my major advisor, Dr Prashant Doshi.

  • Mar 2013

    Project Domain:

    WordPress, HTML, CSS, JQuery, Javascript, PHP, MySQL, Facebook API

    Project Description:

    Website: http://www.phil-athens.com/

  • Mar 2013

    Project Domain:

    ExpressionEngine, HTML, CSS, JQuery, Javascript, PHP

    Project Description:

    Website: http://vet.uga.edu/

  • Mar 2013

    Project Domain:

    Responsive Web Design, ExpressionEngine, HTML, CSS, JS/JQuery

    Project Description:

    Website: http://muthuchandrasekaran.com/

    This started as a responsive web design experiment and I loved the way it looked, so I turned this into my web portfolio which I keep up to date using a content management system called ExpressionEngine.

  • Dec 2012

    Project Domain:

    ExpressionEngine, HTML, CSS, Javascript, JQuery

    Project Description:

    Website: http://dime4dollar.com/

  • Aug 2011

    Project Domain:

    HTML, CSS, PHP, JavaScript/JQuery

    Project Description:

    Website: http://isa.uga.edu/

    I designed and developed the entire website for the Indian Students Association (ISA) as a newly recruited Web Master for the ISA Committee of 2011 – 2012. This was a voluntary project and the website went live on Aug 15, 2011.

  • Jun 2011

    Project Domain:

    HTML, Flash AS3, CSS, PHP, JavaScript/JQuery, Dreamweaver, Photoshop

    Project Description:

    Website: http://www.nit-mut.com/

    This website is now offline. I built and designed this site and all its graphics from scratch as a surprise to my wife for our wedding. This website included functionality like guest book, RSVP, music etc.

  • Feb 2011

    Project Domain:

    ColdFusion, MS SQLServer, Flash AS3, HTML, CSS, Javascript, JQuery

    Project Description:

    Website: http://www.40gallonchallenge.org/

  • Aug 2009

    Project Domain:

    Perl, PHP, MySQL, Google Adwords API, Custom Search API, Dreamweaver

    Project Description:

    Website: http://picobarn.com/

    Consulted for Picobarn. Built a Keyword Analysis Tool and the Picobarn Index - an improvement over the KEI index to measure goodness of keywords based on their search volume and competition.


Contact me

Send me a message





Please enter the word you see in the image :

Last Modified on November 15, 2016