Madhavun Candadai

I am an Artificial Intelligence Research Scientist at Path Robotics where I work on developing algorithms that train robots to perform a variety of manufacturing tasks. I recently received my Ph.D. in Cognitive Science, with a minor in Computer Science in May 2020 from Indiana University, Bloomington under the mentorship of Dr. Eduardo Izquierdo.

Areas of research interest: (Deep) Reinforcement Learning, Computational Neuroscience, Robotics, Evolutionary Robotics, Information Theory, Complex Systems, Dynamical Systems Theory, Artificial Life.

Education and Experience

Artificial Intelligence Research Scientist
Path Robotics

June 2020 - Present


Research Intern
Intel A.I. Labs

Summer 2018


Ph.D., Cognitive Science (minor in Computer Science)
Indiana University, Bloomington.

2015 - 2020

Student Reseacher
Cincinnati Children's Hospital Medical Center

2015

M.S. Electrical Engineering
University of Cincinnati

2015

Associate System Engineer
IBM

2011 - 2012

B.Tech. Electronics and Communication Engineering
Amrita School of Engineering

2011

Publications

See Google Scholar or resume for an updated list of papers, talks and posters.

Computational Neuroscience / Cognitive Science

Candadai, Madhavun, and Eduardo J. Izquierdo. "Sources of predictive information in dynamical neural networks." Scientific reports 10.1 (2020): 1-12. [pdf]
Candadai, Madhavun "Information theoretic analysis of computational models as a tool to understand the neural basis of behaviors" submitted to arxiv [pdf]
Benson, Lauren V., Madhavun Candadai, and Eduardo J. Izquierdo. "Neural reuse in multifunctional neural networks for control tasks." Artificial Life Conference Proceedings. One Rogers Street, Cambridge, MA 02142-1209 USA journals-info@ mit. edu: MIT Press, 2020. [pdf]
Candadai, Madhavun, Matthew Setzler, Eduardo J. Izquierdo, and Tom Froese. "Embodied dyadic interaction increases complexity of neural dynamics: A minimal agent-based simulation model." Frontiers in Psychology 10 (2019): 540. [pdf]
Vasu, Madhavun Candadai, and Eduardo J. Izquierdo. "Multifunctionality in embodied agents: Three levels of neural reuse." arXiv preprint arXiv:1802.03891 (2018). Proceedings of the 40th Cognitive Science Conference, 2018. [pdf]
Vasu, Madhavun Candadai , and Eduardo J. Izquierdo. "Information Bottleneck in Control Tasks with Recurrent Spiking Neural Networks". International Conference on Artificial Neural Networks. (ICANN) Springer, Cham, 2017. [pdf]
Vasu, Madhavun Candadai , and Eduardo J. Izquierdo. "Evolution and Analysis of Embodied Spiking Neural Networks Reveals Task-Specific Clusters of Effective Networks." In Proceedings of Genetic and Evolutionary Computing Conference, pp. 75-82. GECCO, 2017. [pdf]
Nominated for Best Student Paper, 2017, by International Society for Artificial Life: Student Chapter

Machine Learning / A.I.

Leite, Abe, Madhavun Candadai, and Eduardo J. Izquierdo. "Reinforcement learning beyond the Bellman equation: Exploring critic objectives using evolution." Artificial Life Conference Proceedings. One Rogers Street, Cambridge, MA 02142-1209 USA journals-info@ mit. edu: MIT Press, 2020. [pdf]
Todd, Graham, Madhavun Candadai, and Eduardo J. Izquierdo. "Interaction between evolution and learning in nk fitness landscapes." Artificial Life Conference Proceedings. One Rogers Street, Cambridge, MA 02142-1209 USA journals-info@ mit. edu: MIT Press, 2020. [pdf]
Dwiel, Zach*, Madhavun Candadai*, Mariano J. Phielipp. (2019, November). On Training Flexible Robots using Deep Reinforcement Learning. In 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE. (Accepted) [pdf] (*equal contribution)
Dwiel, Zach, Madhavun Candadai, Mariano J. Phielipp, Arjun K. Bansal. "Hierarchical policy learning is sensitive to goal space design". In Task-Agnostic Reinforcement Learning Workshop (TARL), of International Conference on Learning Representations (ICLR), 2019. [pdf]
Candadai, Madhavun, Aashay Vanarase, Mei Mei, and Ali A. Minai. "ANSWER: An unsupervised attractor network method for detecting salient words in text corpora." In International Joint Conference on Neural Networks 2015, pp. 1-8. IEEE, 2015. [pdf]

Other

Candadai, M., & Izquierdo, E. J. (2019). infotheory: A C++/Python package for multivariate information theoretic analysis. arXiv preprint arXiv:1907.02339. [pdf]

Dissertation

Candadai, M. (2020). Bits from Behaviors: Understanding Function Using Information in Embedded, Embodied, and Dynamical Neural Networks (Doctoral dissertation, Indiana University). [pdf]
Candadai Vasu, M. (2015). ANSWER: A Cognitively-Inspired System for the Unsupervised Detection of Semantically Salient Words in Texts (Master's thesis, University of Cincinnati). [pdf]

Open Source Projects

Infotheory

A C++/Python package for information theoretic analysis.