Shiwali Mohan

About

I am a doctoral candidate in Computer Science and Engineering at The University of Michigan Ann Arbor. I have been working with Professor John Laird on the Soar Cognitive Architecture in the Michigan AI Lab for past two years. My work has focused on symbolic agents and how they can learn to navigate new, complex environments. We have looked at agents that start with minimal background knowledge about an environment and build relevant, relational structures using reinforcement learning in the Infinite Mario RL domain. Currently, I am looking at how an agent can learn new, abstract actions when instructed by a human expert. I am also interested in agents that can explain their actions, thus giving a human instructor a deeper understanding of the agent's knowledge and its reasoning process.

I am from India, where I received an undergraduate degree in Instrumentation and Control Engineering from University of Delhi. After four years of struggling with control equations and differential calculus, I ended up loving programming and other geeky, Computer Science-y things. I worked with Yahoo! India as a software engineer on their Strategic Data Solutions team for some time, before deciding to denounce the real world, and the money that comes with it and became a graduate student.

Contact

Email
shiwali[at]umich[dot]edu
shiwali[dot]mohan[at]gmail.com.

Snail-mail
The CSE building,
2260 Hayward Street #3844,
Ann Arbor, MI 48109.

Phone 734-763-0150

Networking
Mendeley | LinkedIn

Research

Publications

2011

Shiwali Mohan, John E. Laird: Towards Situated, Interactive, Instructable Agents in a Cognitive Architecture, to appear In the Proceedings of the AAAI Fall 2011 Symposium on Advances in Cognitive Architectures.
Shiwali Mohan, John E. Laird: An Object-Oriented Approach to Reinforcement Learning in An Action Game, to appear In the Proceedings of AIIDE 2011 (paper)

2010

Shiwali Mohan, John E. Laird: Relational Reinforcement Learning in Infinite Mario (link), AAAI conference on Artificial Intelligence 2010.

2009

Shiwali Mohan, John E. Laird: Learning to Play Mario (link), Technical Report - CCA-TR-2009-03, Ann Arbor MI 2009

2008

Niladri Chatterjee, Shiwali Mohan: Discovering Word Senses from Text Using Random Indexing (link), CICLing 2008: 299-310 (Best Paper Award)

2007

Niladri Chatterjee, Shiwali Mohan: Extraction-Based Single-Document Summarization Using Random Indexing (link), ICTAI 2007: 448-455

Talks and Presentations

June 2011: Modular Reinforcement Learning in Soar (talk), 31st Soar Workshop, Ann Arbor, MI
June 2011: Towards An Architecture for Learning with Instruction (talk), 31st Soar Workshop, Ann Arbor, MI
July 2010: Relational Reinforcement Learning in Infinite Mario (poster), AAAI 2010 Poster Session, Atlanta, Georgia.
May 2010: Learning Background Knowledge through Instruction (poster), Computer Science and Engineering, University of Michigan, Ann Arbor.
May 2010: Learning Background Knowledge through Instruction (link), 30th Soar Workshop, Ann Arbor.
May 2010: Reinforcement Learning in Infinite Mario (link), 30th Soar Workshop, Ann Arbor.
July 2009: Learning to Play Mario (link), at the 29th Soar Workshop, Ann Arbor.

Class Projects

April 2011: On Grounded Language Generation in Intelligent Agents, EECS 595, Natural Language Processing (talk) (paper).
September 2010: Relational Reinforcement Learning in Infinite Mario, Preliminary Examination. (paper)
December 2009: Classification of Executed and Imagined Motor Movement EEG Signals, EECS 545, Machine Learning. (paper)
December 2009: Towards a Resource Aware Scheduler in Hadoop, EECS 589, Advanced Computer Networks. (paper)

Interests

Design

I am dabbling into design and composition, and am interested in variety of things from doodling, ambigrams, type to web design. I am a novice in most of these areas (except doodling maybe), and if you ask me questions, I would respond by scratching my head. However, I will look it up and probably would have an opinion about it. I am also learning how to compose good photographs and I own a cool Nikon D3000.

Dance

I trained in Bharat Natyam when I was a child, and dancing has been a passion since. During college, I was involved in direction, production and choreography with a team that performed stories through dance. I have been dancing on and off for the most of my life and have have picked up tiny bits of Modern and Ballroom, although I have just begun my formal training in western dance styles. I have been learning Jazz and Hip-hop at the Dance Theatre Studio in Ann Arbor for some time now, and I am very interested in fusion styles. Oh, and I totally rock Bollywood.

Links

Reinforcement Learning

Natural Language

  1. SimpleNLG : A Java library to generate correct English sentences.
  2. CLAIR

Author: Shiwali Mohan

Org version 7.5 with Emacs version 23