publications

2024

1. Automated experimental design of safe rampdowns via probabilistic machine learning

   Viraj Mehta, Jayson L Barr , Joseph Abbate , and 5 more authors

   Nuclear Fusion, 2024

   Abs HTML

   Typically the rampdown phase of a shot consists of a decrease in current and injected power and optionally a change in shape, but there is considerable flexibility in the rate, sequencing, and duration of these changes. On the next generation of tokamaks it is essential that this is done safely as the device could be damaged by the stored thermal and electromagnetic energy present in the plasma. This works presents a procedure for automatically choosing experimental rampdown designs to rapidly converge to an effective rampdown trajectory. This procedure uses probabilistic machine learning methods paired with acquisition functions taken from Bayesian Optimization. In a set of 2022 experiments at DIII-D, the rampdown designs produced by our method maintained plasma control down to substantially lower current and energy levels than are typically observed. The actions predicted by the model significantly improved as the model was able to explore over the course of the experimental campaign.

2023

1. Near-optimal Policy Identification in Active Reinforcement Learning

   Xiang Li* , Viraj Mehta*, Johannes Kirschner* , and 5 more authors

   In International Conference on Learning Representations (oral, top 5% of accepted papers) , 2023

   HTML
2. Sample Efficient Reinforcement Learning from Human Feedback via Active Exploration

   Viraj Mehta, Vikramjeet Das , Ojash Neopane , and 4 more authors

   arXiv preprint arXiv:2312.00267, 2023

   HTML

2022

1. An Experimental Design Perspective on Model-Based Reinforcement Learning

   Viraj Mehta, Biswajit Paria , Jeff Schneider , and 2 more authors

   In International Conference on Learning Representations , 2022

   HTML
2. Variational autoencoders in the presence of low-dimensional data: landscape and implicit bias

   Frederic Koehler , Viraj Mehta, Andrej Risteski , and 1 more author

   In International Conference on Learning Representations , 2022

   HTML
3. Sample-efficient Plasma Control by Planning for Optimal Trajectory Information

   Viraj Mehta, Ian Char , Joseph Abbate , and 5 more authors

   In ICML Workshop on Adaptive Experimental Design and Active Learning in the Real World , 2022
4. Exploration via Planning for Information about the Optimal Trajectory

   Viraj Mehta, Ian Char , Joseph Abbate , and 5 more authors

   In Advances in Neural Information Processing Systems , 2022

   HTML

2021

1. Neural Dynamical Systems: Balancing Structure and Flexibility in Physical Prediction

   Viraj Mehta, Ian Char , Willie Neiswanger , and 5 more authors

   In IEEE Conference on Decision and Control , 2021

   HTML
2. Representational aspects of depth and conditioning in normalizing flows

   Frederic Koehler , Viraj Mehta, and Andrej Risteski

   In International Conference on Machine Learning , 2021

   HTML
3. Controlling Plasma Profiles in a Learned Model via Reinforcement Learning

   Viraj Mehta, Joseph Abbate , William Conlin , and 2 more authors

   In American Physical Society Division of Plasma Physics Annual Meeting , Nov 2021

   HTML
4. BATS: Best Action Trajectory Stitching

   Viraj Mehta, Ian Char , Adam Villaflor , and 1 more author

   In NeurIPS Offline Reinforcement Learning Workshop , 2021
5. Bayesian Active Reinforcement Learning

   Viraj Mehta, Biswajit Paria , Jeff Schneider , and 2 more authors

   In NeurIPS Workshop on the Ecological Theory of Reinforcement Learning , 2021

2018

1. DeformNet: Free-Form Deformation Network for 3D Shape Reconstruction from a Single Image

   Andrey Kurenkov , Jingwei Ji , Animesh Garg , and 4 more authors

   In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV) , 2018
2. Learning Task-Oriented Grasping for Tool Manipulation from Simulated Self-Supervision

   Kuan Fang , Yuke Zhu , Animesh Garg , and 4 more authors

   In Proceedings of Robotics: Science and Systems , Jun 2018

2017

1. Towards Grasp Transfer using Grasp Deformation

   Andrey Kurenkov , Viraj Mehta, Jingwei Ji , and 2 more authors

   1st Conference on Robot Learning (short talk), Jun 2017

2013

1. Effects of chemical inhibitors and apyrase enzyme further document a role for apyrases and extracellular ATP in the opening and closing of stomates in Arabidopsis

   Greg Clark , Viraj Mehta, Cameron Darwin , and 4 more authors

   Plant Signaling and Behavior, Jun 2013