Nived Rajaraman

Electrical Engineering and Computer Science Department, University of California, Berkeley 264 Cory Hall Berkeley, CA 94720-1770 Email: nived [dot] rajaraman [at] berkeley [dot] edu

About me

I am a fourth year Ph.D. student at UC Berkeley, jointly advised by Jiantao Jiao and Kannan Ramchandran. I am affiliated with the BLISS and BAIR labs.

I work on a variety of topics in machine learning with a focus on the statistical and computational aspects of adaptive decision making problems and reinforcement learning. I am also interested in nonconvex optimization and federated learning. I appreciate papers with well motivated theoretical formulations which either explain curious practical phenomena or ultimately provide intuitions missing in existing practical approaches.

I've had the great fortune of working with two amazing undergrads, Alex Deweese and Matt Peng. Alex and I had been thinking about algorithms for online RL with a focus on achieving optimal sample complexity guarantees. Matt and I worked on implementing performant algorithms for Imitation Learning based on a novel technique known as Replay Estimation.

I organize the BLISS seminar and CLIMB seminar at Berkeley. Shoot me an email if you are interested in giving a talk at either venue!

In a previous life, I was a dual degree student at the Department of Electrical Engineering, IIT Madras. I am fortunate to have had Ravishankar Krishnaswamy and Prof. Andrew Thangaraj as thesis advisors and work closely with Prof. Rahul Vaze.

I will be participating in the AIDS LifeCycle 2023. You can support me by clicking here. Every little bit counts!

A full list of publications can be accessed here.

Select recent publications

1. On the Value of Interaction and Function Approximation in Imitation Learning (NeurIPS 2021)

With Jingbo Liu, Yanjun Han, Lin F. Yang, Jiantao Jiao and Kannan Ramchandran

2. Toward the Fundamental Limits of Imitation Learning (NeurIPS 2020) (ArXiv) (video)

With Lin F. Yang, Jiantao Jiao and Kannan Ramchandran

3. Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally (ArXiv)

With Yanjun Han, Lin F. Yang, Kannan Ramchandran and Jiantao Jiao

4. Semi-supervised Active Regression (ArXiv)

With Devvrit and Pranjal Awasthi

5. FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning (ArXiv)

With Swanand Kadhe, O. Ozan Koyluoglu and Kannan Ramchandran