I am a PhD candidate at MIT in Josh Tenenbaum’s group. I am also affiliated with the MIT Probabilistic Computing Project, MIT CSAIL and Brain and Cognitive Sciences (BCS).

I want to build and understand intelligence. I am particularly interested in building agents that autonomously learn to solve many goals by simply interacting with the environment. To achieve this, I use the following methodologies in my research: Deep Reinforcement Learning, Deep Learning and Probabilistic Programming.

Github: https://github.com/mrkulk

Email: tejask@mit.edu | tejasdkulkarni@gmail.com

Organizations: NIPS/ICML/ICLR Reviewer, NIPS Blackbox Learning Workshop organizer

  • Present2012

    PhD Candidate

    Massachusetts Institute of Technology, Cambridge

  • Aug '15May '15

    Research Internship

    Google Deepmind

  • 20102007

    BS in Electrical and Computer Engineering

    Purdue University

Recent Papers

Deep Successor Reinforcement Learning
Tejas Kulkarni*,  Ardavan Saeedi*, Simanta Gautam, Sam Gershman
arXiv:1606.02396 [paper] [code] * – Equal contribution

 

Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation
Tejas Kulkarni*,  Karthik Narasimhan*, Ardavan Saeedi, Joshua Tenenbaum
ICML 2016, Workshop on Temporal Abstractions in RL. (Oral Talk[paper] * – Equal contribution

 

Understanding Visual Concepts with Continuation Learning
William Whitney, Michael Chang, Tejas Kulkarni, Joshua Tenenbaum
ICLR 2016 Workshop [paper] [project webpage]

 

Deep Convolutional Inverse Graphics Network
Tejas Kulkarni*, William Whitney*, Pushmeet Kohli, Joshua Tenenbaum
NIPS 2015 (Oral Presentation) [paper] [project webpage]  *-Equal contribution

 

Picture:a probabilistic programming language for scene perception
Tejas Kulkarni*, Pushmeet Kohli, Joshua Tenenbaum, Vikash Mansinghka
CVPR 2015 (Best Paper Hon’ Mention Award) [paper] [project webpage] [PressMIT News, Phys.org, PCWorld, KurzweilAI, Motherboard]

 

Language understanding for text-based games using deep reinforcement learning
Karthik Narasimhan*, Tejas Kulkarni*, Regina Barzilay
EMNLP 2015 (Best Paper Hon’ Mention Award) [paper] [code] [Press: MIT News] *-Equal contribution

 

Efficient analysis-by-synthesis in vision: A computational framework, behavioral tests, and comparison with neural representations
Ilker Yildrim, Tejas Kulkarni, Winrich Friewald, Joshua Tenenbaum
Proceedings of the Thirty-Seventh Annual Conference of the Cognitive Science Society 2015 [paper

 

Variational Particle Approximations
Ardavan Saeedi*, Tejas Kulkarni*, Vikash Mansinghka, Sam Gershman
NIPS Workshop on Advances in Variational Inference 2015 (Oral Presentation) [paper*-Equal contribution

 

Deep Generative Vision as Approximate Bayesian Computation
Tejas Kulkarni, Ilker Yildirim, Pushmeet Kohli, Winrich Freiwald, Joshua  Tenenbaum
NIPS Workshop on Approximate Bayesian Computation 2014 [paper]

 

Approximate bayesian image interpretation using generative probabilistic graphics programs
Vikash Mansinghka*, Tejas Kulkarni*, Yura Perov, Joshua Tenenbaum
NIPS 2013 (Oral Presentation) [paper*-Equal contribution

Grateful for the following recognitions

  • Best Paper Honorable Mention Award @ EMNLP 2015
  • Best Paper Honorable Mention Award @ IEEE CVPR 2015
  • MIT Center for Brains, Machines, and Minds Fellowship
  • Leventhal Fellowship
  • Henry H. Singleton Fellowship

Recent Invited Talks

  • Boston Machine Learning Meetup, 2015
  • NIPS, Montreal, 2015
  • Oxford University, 2015
  • CVPR, Boston, MA, USA, 2015
  • Center for Brain Machines and Minds, MIT, Cambridge, 2014