Research & Academic Projects
Research Projects
Communicating with Computers (DARPA project)
Type : Research
Collaborators: Sriraam Natarajan, Dan Roth, Martha Palmer, Julia Hockenmaier, Janna Doppa
Description: The aim of this DARPA funded project is to develop effective communication strategies involving humans-in-the-loop. Particularly, we focus on effective learning and planning techniques that can learn from and possibly teach a human to jointly solve problems.
Status : Ongoing.....
Click to read more about CwC ...
Publications :
Mayukh Das, Phillip Odom, Md. Rakibul Islam, Jana Doppa, Dan Roth, and Sriraam Natarajan, Preference-Guided Planning: An Active Elicitation Approach, International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2018.
Anjali Narayan-Chen, Colin Graber, Mayukh Das, Md Rakibul Islam, Soham Dan, Sriraam Natarajan, Janardhan Rao Doppa, Julia Hockenmaier, Martha Palmer, Dan Roth, Towards Problem Solving Agents that Communicate and Learn, Proceedings of the First Workshop on Language Grounding for Robotics @ ACL (RoboNLP), 2017.
Mayukh Das, Md. Rakibul Islam, Jana Doppa, Dan Roth, and Sriraam Natarajan, Active Preference Elicitation for Planning, AAAI Workshop on Human-Machine Collaborative Learning (HMCL), 2017.
Scaling in Probabilistic Logic Models via Graph & Hypergraph DB
Type : Research
Collaborators: Sriraam Natarajan, Yuqing Wu (Pomona College), Kristian Kersting (TU - Dortmund), Tushar Khot (AI2), Gautam Kunapuli
Description : Over the past decade, exploiting relations and symmetries within probabilistic models has been proven to be surprisingly effective at solving large scale problems. One of the key operations inside these lifted approaches is counting - be it for parameter/structure learning or for efficient inference. Counting satisfied instances of generalized structural features is hard and becomes a major in scalability. We investigate and propose generalized approximate counting techniques for efficient learning and inference leveraging the power of graph structured databases.
Status: Ongoing
Publications :
Mayukh Das, Devendra Singh Dhami, Gautam Kunapuli, Kristian Kersting, and Sriraam Natarajan, Fast Relational Probabilistic Inference and Learning: Approximate Counting via Hypergraphs, The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI), 2019.
Mayukh Das, Yuqing Wu, Tushar Khot, Kristian Kersting and Sriraam Natarajan, Scaling Lifted Probabilistic Inference and Learning Via Graph Databases, SIAM International Conference on Data Mining (SDM), 2016.
Mayukh Das, Yuqing Wu, Tushar Khot, Kristian Kersting, and Sriraam Natarajan, Graph-based Approximate Counting for Relational Probabilistic Models, International Workshop on Statistical Relational AI (StarAI) 2015.(PDF)
Course Projects
Malware Classification System
Type: Course Project [for CSCI-B565 Data Mining] @ IUB
Technologies: Java, Python, Scikit-learn package
Description: Multi-classclassificationofmalwaresbasedonsourcecodeandassemblylevelinstructionsofexecutables. Information retrieval / Feature Engineering + Machine Learning).
- Feature construction/extraction from malware source code + and assembly level instruction files using JAVA
- Engineered features are used to build classifiers using off-the-shelf systems scikit-learn (python).
We showed that with reasonable feature construction (Data cleaning followed by unigram, bigram and N-gram extractions), Random Forests perform superbly with and gives a test accuracy of greater that 98%. To get code and the final report please go to (my Github).
Tech Supplies Requisition System
Type: Course Project [for CSCI-B561 Advanced Database Concepts] @ IUB
Technologies: PHP, MySQL, HTML, CSS and Javascript.
Description: This is a Web Application. This is a system which automates the process of raising and managing requests for technical utility commodities in an office environment. Also has auto-generated reports on usage data for monitoring purposes. The application essentially has 3 modules, the Requisition module (for raising, viewing and tracking requests), the Inventory module (for managing and tracking inventory movement of the tech utility articles) and the Reporting module (for monitoring the usage). Usage monitoring adds value in such a model because the commodities are given to the employees or students and over-usage can be a prominent issue.
Below are some screenshots of the web application. To visit the online website please click here.
Mobile Application for online voting
Type: Undergraduate Final year project [Credits to: Debargha Ganguly, Anirban Acharjee, Shauvik Roy, Sumit Roy and Mayukh Das]
Technologies: J2ME, HTML, CSS, Oracle Database, SQL.
Description: It is an Java based mobile application with which people can vote for any competition or event. Users need to install the client into their mobile and the server will be in another mobile device. Users will send SMS to the phone number of the server. The standard SMS gateway is bypassed hence SMS cost will not be charged from the user. We have used J2ME, hence the mobile phones must be J2ME enabled. We were a team of 5 members. Along with developing the J2ME code my additional role was to configure and build the web-server and the database underneath it and integrate it with the mobile device that was used as the SMS server.The Webserver and the database was used as backend storage for the votes received and thereafter for the calculation of the poll results. The votes were forwarded from the mobile device to the web-server using HttpPost.
