Daniel Russo
@DanielRuss0
Followers
1K
Following
18
Media
7
Statuses
89
Researcher. Prof of OR at Columbia. Tweeting about reinforcement learning.
New York, NY
Joined May 2020
🚨New workshop alert 🚨 Calling all RL researchers in the New York area 🧠🗽 Present your work at the first-ever New York Reinforcement Learning Workshop (NYRL), co-organized by Amazon, Columbia Business School & NYU Tandon School of Engineering. https://t.co/v7MropcLcz
2
15
57
Find more info here: https://t.co/VNCy9NZ3xe The initiative is led by my colleagues Hong Namkoong, Ciamac Moallemi, Tianyi Peng, and I.
1
0
2
There are multiple postdoc positions available as part of an exciting new AI-agent initiative at Columbia that tackles challenges at the frontier of agentic systems and sequential decision-making. I am not very active here so please help me spread the word!
1
19
56
I'm speaking at this ML symposium at the New York Academy of Sciences next Friday (Oct 18). Come by and say hello.
2
0
5
@DeployableRL workshop is happening this Friday at @RL_Conference. I will be moderating a panel hosting Minmin Chen, Omer Gottsman, @DanielRuss0, Kaushik Subramanian and Cathy Wu. Please vote here https://t.co/g9cWkH04qQ to push your favorite production RL question to the top!
docs.google.com
Help us prioritize panel questions at the Deployable RL Workshop at RLC 2024. Workshop website: https://deployable-rl.github.io/ Panelists: Minmin Chen Omer Gottsman Daniel Russo Kaushik Subramanian...
0
6
11
Excited to give a talk here this Friday. Come by if you're in the Boston area.
Together with @eshedob, we are organizing a Reinforcement Learning Day at Boston University on May 10th: https://t.co/J16ZDNY6hD We have a great lineup of speakers who will talk about recent progress in RL. Registration & attendance is free: please share and consider
1
1
21
This book is a timeless gem. The thing to read to get a deeper grasp of duality.
There is a now obscure book everybody interested in optimization should read; we call it the Red Book. The real title is Optimization by Vector Space Methods, by Luenberger. It’s really about convex duality and functional analysis. It’s lovely and how I learned duality.
0
0
9
(2/2) A team of 100 people manually recorded transaction data on punch cards (the state), which were fed into a specialized machine (the policy) that decided whether to print a mailing labels for a specific catalog. If only he could have seen RL at a modern recommender system!
0
0
4
(1/2) Cool article commenting on the use of MDPs to optimize catalog mailing policies at Sears as early as 1960. https://t.co/HYTs1hyuJv Commentary is from Ron Howard, inventor of policy iteration. It was first published in 1978.
1
4
38
Are you a PhD student interested in tackling real-world problems using ML? My team (@SpotifyResearch) is hiring interns for next summer.
3
29
167
Applying for PhD programs? Please consider the program at the Decision, Risk, and Operations division of Columbia. See flyer:
0
0
16
Check out the poster in the rl4reallife workshop to learn about how this style of thinking has impacted podcast recommendations. (Full paper on that is forthcoming) https://t.co/cX8pgvIBxk
sites.google.com
Website @ NeurIPS 2022 (videos, posters, etc.)
0
0
2
Check out our neurips paper and blog post on synthesizing ideas from reinforcement learning and survival analysis. https://t.co/pjdLsBjAUV
research.atspotify.com
Survival Analysis Meets Reinforcement Learning | Spotify Research
1
0
4
Spotify has been investing in reinforcement learning. Check out some work with the brilliant @lucasmaystre at Neurips 2022.
We decided to invest in reinforcement learning as we believe that personalization should deliver a lifetime of content, rather than optimize for clicks. This #NeurIPS2022 paper from colleagues @lucasmaystre and @DanielRuss0 describes some of the work we are doing towards this.
2
3
67
Thank you to the prize committee! It's quite humbling to be recognized alongside Prof. Nesterov and past winners. Getting to hangout and celebrate with Ben (for the first time since Covid) made this especially cool.
Congratulations to Ben Van Roy and @DanielRuss0, the recent winners of this year’s Lanchester Prize for their work on randomised exploration schemes. Their research has helped form a foundation for exploration methods in #RL and has had a major impact in the AI field.
8
2
94
Any retweets would be highly appreciated. Especially if you're outside my usual operations research bubble (e.g. in CS or Econ).
0
0
0
My department, the Decision, Risk, and Operations Division of Columbia Business School, has an opening for tenure-track appointments at the assistant or associate professor level. Please share this link with anyone who many be interested! https://t.co/eCVYiNvYyd
1
3
15
The updated paper has a second example, designed to break a modified variant of API which is basically PG with large stepsizes. To get the robustness benefits of PG, you need small stepsizes (it needs to act like a gradient method, not like policy iteration)
0
0
1