Devendra Singh Chaplot

I am a Ph.D. student at the Machine Learning Department, School of Computer Science, Carnegie Mellon University. I work with Prof. Ruslan Salakhutdinov on Deep Reinforcement Learning and its applications in Robotics and NLP. Prior to joining CMU, I graduated from IIT Bombay, India with a B.Tech in Computer Science & Engineering and a minor in Applied Statistics in August 2014, where I worked with Prof. Pushpak Bhattacharyya. I also worked for about a year at Samsung Electronics HQ in South Korea.

CV / Google Scholar / Github / Twitter
Email: chaplot[at]cs.cmu.edu, Office: GHC 8208

Upcoming

27 Jun - 30 Jun '18
Oral presentation for our paper on Learning Cognitive Models using Neural Networks at AIED-18.
21 May - 25 Aug '18
Excited to intern with Dhruv Batra and Devi Parikh at Facebook AI Research, Menlo Park.

Updates

Jun '18
New AIED-18 paper on Learning Cognitive Models using Neural Networks posted on arxiv.
[PDF]
Jun '18
Paper on end-to-end global pose estimation using Neural Graph Optimization received Best Paper Award at the CVPR-18 Deep Learning for Visual SLAM workshop.
[PDF]
Jun '18
New ICML-18 paper on Gated Path Planning Networks posted on arxiv.
[PDF] [Code]
Apr '18
Released code, environment and pre-trained models for ICLR-18 Localization paper.
Mar '18
Gave a talk on DeepRL in 3D environments at the Nvidia GTC 2018.
Feb '18
New paper on our work at Apple AI Research on end-to-end global pose estimation using Neural Graph Optimization posted on arxiv.
[PDF]
Feb '18
Gave two talks at AAAI-18, on WSD and Language Grounding.
Jan '18
Paper on Active Neural Localization accepted at ICLR-18.
Jan '18
Released code, environment and pre-trained model for AAAI-18 Language Grounding paper.
Jan '18
AAAI-18 WSD paper available on arXiv.
[PDF]
Dec '17
Released code and pre-trained models for our Doom AI agent, Arnold. Play against our Doom AI agent which won the Visual Doom AI Competition 2017 Full Deathmatch.
Dec '17
Gave a talk on Active Neural Localization at the NIPS-17 DeepRL Symposium.
Nov '17
Two papers accepted for oral presentation at AAAI-18 on WSD and Language Grounding.
Sep '17
Our Doom AI agent, Arnold won the Visual Doom AI Competition 2017 Full Deathmatch.
Jun '17
MIT TechReview and Inverse write about our recent paper on language grounding!

Select Research Projects

Learning to localize

Active Neural Localization
Devendra Singh Chaplot, Emilio Parisotto, Ruslan Salakhutdinov. (2018)
6th International Conference on Learning Representations (ICLR-18), Vancouver, Canada
PDF / Code / Environment / Pre-trained models / Demo videos / Talk / Slides

Learning to execute natural language instructions

Gated-Attention Architectures for Task-Oriented Language Grounding
Devendra Singh Chaplot, Kanthashree Mysore Sathyendra, Rama Kumar Pasumarthi, Dheeraj Rajagopal, Ruslan Salakhutdinov. (2018)
32nd AAAI Conference on Artificial Intelligence (AAAI-18), New Orleans, USA.
PDF / Code / Environment / Pre-trained models / Demo videos
Media: MIT TechReview, Inverse

Learning to play deathmatches in Doom

Playing FPS Games with Deep Reinforcement Learning
Guillaume Lample*, Devendra Singh Chaplot*. (2017)
31st AAAI Conference on Artificial Intelligence (AAAI-17), San Francisco, USA.

Arnold: An Autonomous Agent to play FPS Games (Best Demo Award)
Devendra Singh Chaplot*, Guillaume Lample*. (2017)
31st AAAI Conference on Artificial Intelligence (AAAI-17), San Francisco, USA. (demo)

PDF / Code / Pre-trained models / Demo videos (300K+ views)
Media: TechCrunch, Popular Science, Engadget, Daily Mail, Salon, Kotaku, ScienceAlert, Pittsburgh Post-Gazette, Inverse, CMU News
First place Visual Doom AI Competition 2017 Full Deathmatch
Second place Visual Doom AI Competition 2016 Full Deathmatch

Publications

Gated Path Planning Networks

Lisa Lee, Emilio Parisotto, Devendra Singh Chaplot, Eric Xing, Ruslan Salakhutdinov. (2018)
35th International Conference on Machine Learning (ICML-18), Stockholm, Sweden
PDF, Code

Learning Cognitive Models using Neural Networks (oral)

Devendra Singh Chaplot, Christopher MacLellan, Ruslan Salakhutdinov, Kenneth Koedinger. (2018)
19th International Conference on Artificial Intelligence in Education (AIED-18), London, UK
PDF

Global Pose Estimation with an Attention-based Recurrent Network (oral, Best Paper Award)

Emilio Parisotto*, Devendra Singh Chaplot*, Jian Zhang, Ruslan Salakhutdinov. (2018)
Conference on Computer Vision and Pattern Recognition (CVPR-18), Deep Learning for Visual SLAM workshop, Salt Lake City, USA.
PDF

Active Neural Localization

Devendra Singh Chaplot, Emilio Parisotto, Ruslan Salakhutdinov. (2018)
6th International Conference on Learning Representations (ICLR-18), Vancouver, Canada
Also presented as contributed talk at the NIPS-17 Deep RL Symposium, Long Beach, USA
PDF, Code, Environment, Pre-trained models, Demo videos, Talk, Slides

Gated-Attention Architectures for Task-Oriented Language Grounding (oral)

Devendra Singh Chaplot, Kanthashree Mysore Sathyendra, Rama Kumar Pasumarthi, Dheeraj Rajagopal, Ruslan Salakhutdinov. (2018)
32nd AAAI Conference on Artificial Intelligence (AAAI-18), New Orleans, USA.
PDF, Code, Environment, Pre-trained Models, Demo videos
Media: MIT TechReview, Inverse

Knowledge-based Word Sense Disambiguation using Topic Models (oral)

Devendra Singh Chaplot, Ruslan Salakhutdinov. (2018)
32nd AAAI Conference on Artificial Intelligence (AAAI-18), New Orleans, USA.
PDF

Playing FPS Games with Deep Reinforcement Learning

Guillaume Lample*, Devendra Singh Chaplot*. (2017)
31st AAAI Conference on Artificial Intelligence (AAAI-17), San Francisco, USA.
PDF, Code, Environment, Pre-trained models, Demo videos (300K+ views)
Media: TechCrunch, Popular Science, Daily Mail, Salon, Kotaku, ScienceAlert, Pittsburgh Post-Gazette

Arnold: An Autonomous Agent to play FPS Games (Best Demo Award)

Devendra Singh Chaplot*, Guillaume Lample*. (2017)
31st AAAI Conference on Artificial Intelligence (AAAI-17), San Francisco, USA. (demo)
PDF, Code, Environment, Pre-trained models, Demo videos (300K+ views)
Media: Engadget, CMU News, Inverse

Transfer Deep Reinforcement Learning in 3D Environments: An Empirical Study

Devendra Singh Chaplot, Guillaume Lample, Kanthashree Mysore Sathyendra, Ruslan Salakhutdinov (2016).
30th Annual Conference on Neural Information Processing Systems (NIPS-16), Deep RL Workshop, Barcelona, Spain.
PDF

Unsupervised Word Sense Disambiguation using Markov Random Field and Dependency Parser

Devendra Singh Chaplot, Pushpak Bhattacharyya, Ashwin Paranjape. (2015)
29th AAAI Conference on Artificial Intelligence (AAAI-15), Austin, USA.
PDF, Demo

Data-driven Automated Induction of Prerequisite Structure Graphs

Devendra Singh Chaplot, Yiming Yang, Jaime Carbonell, Kenneth Koedinger. (2016)
9th International Conference on Educational Data Mining (EDM-16), Raleigh, USA.
PDF

Personalized Adaptive Learning using Neural Networks

Devendra Singh Chaplot, Eunhee Rhim, Jihie Kim. (2016)
3rd ACM Conference on Learning at Scale (L@S-16), Edinburgh, UK.
PDF

Predicting Student Attrition in MOOCs using Sentiment Analysis and Neural Networks

Devendra Singh Chaplot, Eunhee Rhim, Jihie Kim. (2015)
17th International Conference on Artificial Intelligence in Education (AIED-15), ISLG Workshop, Madrid, Spain.
PDF

SAP: Student Attrition Predictor

Devendra Singh Chaplot, Eunhee Rhim, Jihie Kim. (2015)
8th International Conference on Educational Data Mining (EDM-15), Madrid, Spain. (demo)
PDF

IndoWordnet Visualizer: A Graphical User Interface for Browsing and Exploring Wordnets of Indian Languages

Devendra Singh Chaplot, Sudha Bhingardive, Pushpak Bhattacharyya. (2014)
7th Global WordNet Conference (GWC-14), Tartu, Estonia.
PDF, Demo

Comparing Model Comparison Methods

Holger Schultheis, Ankit Singhaniya, Devendra Singh Chaplot. (2013)
35th annual conference of the Cognitive Science Society (CogSci-13), Berlin, Germany.
PDF

(* Equal Contribution)

Talks

Playing FPS Games with Deep Reinforcement Learning

Nvidia GTC 2018
Video, Slides

Doom and Unreal Game Engines

Embodied Agents and Environments Workshop, Facebook AI Research, Menlo Park
Slides

Gated-Attention Architectures for Task-Oriented Language Grounding

AAAI 2018, oral
Slides

Knowledge-based Word Sense Disambiguation using Topic Models

AAAI 2018, oral
Slides

Active Neural Localization

NIPS 2017, Deep Reinforcement Learning Symposium
Video, Slides

Demo Videos

Learning to localize

Unreal Environment
Doom Environment


Learning to execute natural language instructions

Different stages of training
Multitask Learning
Zero-shot Learning


Learning to play deathmatches in Doom

Different stages of training
Navigation on known maps
Navigation on unknown maps
Limited Deathmatch on a known map
Full Deathmatch on unknown maps