CS PhD Student in CMU. Interested in reinforcement learning
Pittsburgh, PA
Joined May 2019
- Tweets123
- Following180
- Followers1.1K
- Likes234
Ruosong Wang retweeted
Is a good representation sufficient for sample-efficient offline RL? In practice, offline RL can be very sensitive to distribution shift even when using features from pretrained neural networks on the same task
arxiv.org/abs/2103.04947
w/t @RuosongW, Wu & @ShamKakade6 #icml2021
Ruosong Wang retweeted
More bad news from those mean theoreticians: Unlike in planning with a simulator, even a positive action gap is insufficient for efficient online RL if all one has is linear q^* realizability. Yuanhao Wang's talk is next Tuesday at the RL Theory Seminar. tinyurl.com/rmax5y74
Ruosong Wang retweeted
Our next talk:
06/08: Yuanhao Wang (Princeton, @yuanhaowang3)
"An Exponential Lower Bound for Linearly-Realizable MDPs with Constant Suboptimality Gap"
For details, please see the website:
sites.google.com/view/rltheo…
Ruosong Wang retweeted
Our next talk:
03/09: Ruosong Wang (CMU)
"What are the statistical limits of offline RL with linear function approximation?"
For details, please see the website:
sites.google.com/view/rltheo…
Ruosong Wang retweeted
Advisor abuse of power is a serious problem in academia. I know victims who are reluctant to discuss it for fear of damaging their careers, especially if they are on a visa. I'm really happy to see this MIT org pushing to do something about it. Hopefully others follow.
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
Ruosong Wang retweeted
Alekh Agarwal, Akshay Krishnamurthy, and I finished a major tutorial on "Theoretical Foundations of Reinforcement Learning" (hunch.net/~tforl/) for FOCS (focs2020.cs.duke.edu/program…), but potentially of much broader interest. We'll be doing Q&A Friday.
Ruosong Wang retweeted
Planning with Submodular Objective Functions: Instead of maximizing cumulative reward, maximize the objective induced by submodular function. Standard planning & submodular maximization can be viewed as special cases
arxiv.org/abs/2010.11863
w/ @RuosongW, Zhang, @dchaplot et al
This looks impressive
Introducing SAM: An easy-to-use algorithm derived by connecting PAC Bayesian bounds and geometry of the loss landscape. Achieves SOTA on benchmark image tasks (0.3% error on CIFAR10, 3.9% on CIFAR100) and drastically improves label noise robustness.
arxiv.org/abs/2010.01412
Ruosong Wang retweeted
Duke Comp Sci PhD candidate Hanrui Zhang receives best student paper award at 28th European Symposium on Algorithms (ESA) Sept. 7-9, 2020. A 4th-yr PhD student, Zhang's research area is Econ. & Computation; he's advised by Prof. Vincent Conitzer. Congrats! bit.ly/dukecs-ESA2020-best-s…
Ruosong Wang retweeted
The one by Jelani Nelson (@minilek) on Sketching and Streaming algorithms should be great! 😍 #NeurIPS2020
Thanks to everyone who submitted tutorial proposals for this year's @NeurIPSConf. It was amazing to see all the great work that went in to putting together these proposals. We are happy to announce this year's tutorials at #neurips2020 medium.com/@NeurIPSConf/tuto…
How to explore an environment without the guidance of any reward function? In our recent work (arxiv.org/abs/2006.11274), we use UCB as the bonus to efficiently explore an unknown environment and find good policies for all possible reward functions in the linear MDP setting. 1/4
Our results imply exponential separations on the sample complexity between (1) the linear MDP assumption and the linear Q* assumption, (2) standard RL and reward-free exploration, and (3) reward-free exploration in the generative model and that in the standard RL model. 3/4
Ruosong Wang retweeted
Thank you so much to the awards committee! Also a huge thanks to the past and current ICML chairs and organizers for all their great work for our community! 👍 It is an honor to receive this 😀😀 with such wonderful co-authors: @arkrause, Matthias, and Niranjan!
We are very pleased to announce that the #icml2020 Test of Time award goes to
Gaussian Process Optimization in the Bandit Setting: No Regret and Experimental Design
by Niranjan Srinivas, Andreas Krause, Sham Kakade and Matthias Seeger
icml.cc/Conferences/2020/Awa…
>
Ruosong Wang retweeted
We just released document V1.0 about the #NeurIPS2020 generalization competition to cover competition description (datasets, metrics, etc.), timeline, rules and protocols.
We just released the first document of the competition on our website sites.google.com/corp/view/p…. It covers competition description (datasets, metrics, etc.), timeline, rules and protocols. Hurry up... Phase 1 starts on July 15.
Ruosong Wang retweeted
Do you have a theory or a hypothesis that can predict generalization performance in deep learning? Now, for the first time, there is a competition for it & you can validate your idea @NeurIPSConf #NeurIPS2020. We hope this will instigate progress in understanding generalization.
Ruosong Wang retweeted
Our model (UCLA, covid19.uclaml.org) is one of the 12 models used by @CDCgov to forecast the #COVID19 deaths in the US.
CDC tracks 12 different forecasting models of possible #COVID19 deaths in the US. As of May 11, all forecast an increase in deaths in the coming weeks and a cumulative total exceeding 100,000 by June 1. See national & state forecasts: bit.ly/3cKQIl4.
Our result shows that the sample complexity of RL depends only logarithmically on the planning horizon, even if rewards are correctly normalized. Main technical idea: an eps-net for near-optimal policies and a trajectory simulator which efficiently evaluates a set of policies.
1/ Playing the long game: Is long horizon RL harder than short horizon RL? Clearly, H length episodes scale linearly with H, but counting learning complexity by # episodes rather than # samples accounts for this. So is it any harder? arxiv.org/abs/2005.00527
We are presenting our work on necessary conditions for efficient RL with function approximation at #ICLR2020. Come and chat with us at iclr.cc/virtual/poster_r1gen….