Teodor Mihai Moldovan

Teodor Mihai Moldovan

Currently I am a fifth year Ph.D. student in the EECS department at UC Berkeley where I am co-advised by Prof. Pieter Abbeel and Prof. Michael I. Jordan, working in collaboration with the AMPLab and the UC Berkeley Robotics group.

I obtained my Masters's (MA) degree in Statistics in 2012 from UC Berkeley and my Bachelor of Science (ScB) degree with triple concentration (computer science, physics, mathematics) in 2009 from Brown University, where my research adviser was Prof. Michael J. Black.

moldovan AT berkeley.edu

Current research interests

When learning to control dynamical systems, such as a remote-control helicopter or a bicycle, we intuitively construct a mental model for how the system works, we plan under this uncertain model to explore the capabilities of the system while avoiding crashes. My research focuses on planning under uncertainty in continuous state space using sample-efficient and safe exploration methods that require minimal supervision and system-specific knowledge. I approach the problem using various tools from reinforcement learning, non-parametric Bayesian modelling and optimization.

Dirichlet Process Reinforcement Learning,
Teodor Mihai Moldovan, Michael I Jordan, and Pieter Abbeel. Presented at the 1st Multidisciplinary Conference on Reinforcement Learning and Decision Making, 2013.
poster | slides | pdf | http ]


Safe Exploration in Markov Decision Processes,
Teodor Mihai Moldovan and Pieter Abbeel. In Proceedings of the 29th International Conference on Machine Learning, 2012.
bib | arXiv | poster | slides | pdf | Abstract ]

Risk Aversion in Markov Decision Processes via Near-Optimal Chernoff Bounds,
Teodor Mihai Moldovan and Pieter Abbeel. In Advances in Neural Information Processing Systems 25, 2012.
bib | poster | pdf | Abstract ]

Scaling the mobile millennium system in the cloud,
Timothy Hunter, Teodor Moldovan, Matei Zaharia, Samy Merzgui, Justin Ma, Michael J. Franklin, Pieter Abbeel, and Alexandre M. Bayen. In Proceedings of the 2nd ACM Symposium on Cloud Computing, 2011.
bib | pdf | http | Abstract ]

Denoising archival films using a learned bayesian model,
Teodor Mihai Moldovan, Stefan Roth, and Michael J. Black. In Proceedings of the International Conference on Image Processing, 2006.
bib | pdf | http | Abstract ]


dpcluster ] Variational Dirichlet process clustering for Python. Batch and on-line versions available.