Sridhar Mahadevan

 

Adjunct Professor

Director, Adobe Research




AAAI Fellow

Co-Director, Autonomous Learning Laboratory



Research Interests




Artificial Intelligence

Imagination Science

Machine Learning

Variational Inequalities






CONTACT INFO:




firstname.i.lastname AT gmail

Administrative Assistant: Sarah Byrne

 
 
 

After 17 years of being a full-time faculty member at UMass (from 2001-2018), I am now an Adjunct Professor (effective 9/1/2018). I am also currently a Director at Adobe Research in San Jose, CA.


I am developing a new framework called imagination science. Traditional data science, like statistics, constructs descriptive summaries of experience, and answers the question "What is?", converting experience into probability distributions. Imagination Science explores a much broader palette of questions, including answering "What if?" interventional queries, as well as "Why?" exlanations, and counterfactual imagination-type reasoning. I have given tutorials on this framework at various AI/ML conferences over the past two years, including ECML 2019, IJCAI 2018 and AAAI 2019. I also co-organized a AAAI 2019 Spring Symposium on this topic. We have a new ICML 2020 paper that relates to the topic of imagination, as it addresses how to extend reinforcement learning to nonstationary environments (https://arxiv.org/abs/2005.08158).

Elias Bareinboim, Sridhar Mahadevan, Judea Pearl, Bernhard Scholkopf, and Prasad Tadepalli, Beyond Curve Fitting: Causation, Counterfactuals, and Imagination-based AI, AAAI Spring Symposium, March 25-27, Stanford University, Palo Alto, CA, 2019.

Machines Who Imagine: Beyond Data Science, Tutorial at ECML Conference, September 16 2019. Wurzburg, Germany.

Tutorial on a new framework for ML and AI based on variational inequalities, AAAI 2015, January 26 2015.