Seungjae Ryan Lee

@seungjaeryanlee

Math major in Princeton. | Google Summer of Code '19 w/ TensorFlow. | Research Intern at SK T-Brain

Vrijeme pridruživanja: lipanj 2017.

Tweetovi

Blokirali ste korisnika/cu @seungjaeryanlee

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @seungjaeryanlee

  1. 10. sij

    My first paper got published! But I still need 1-2 ML papers published before I apply to graduate school😅

    Poništi
  2. proslijedio/la je Tweet
    25. pro 2019.
    Prikaži ovu nit
    Poništi
  3. Great papers accepted to ICLR 2020! Here's a summary of the two I particularly enjoyed reading:

    Poništi
  4. proslijedio/la je Tweet
    20. pro 2019.

    What novel techniques in deep RL fundamentally advanced the state of the art this year, in ways that are very likely to be attributable to the novel techniques and not other features of the codebase / optimization tricks / "tuning" by hyperparameter or architecture fiddling?

    Poništi
  5. Poništi
  6. A set of environments with procedurally generated content (level layout, game assets, entity spawn location and timing, etc.) to encourage agents robust to variations. By , , Jacob Hilton, and John Schulman

    Prikaži ovu nit
    Poništi
  7. It is possible capture credit assignment explicitly with "hindsight distribution" and use it to estimate value function and policy gradient. By et al.

    Prikaži ovu nit
    Poništi
  8. Overparametrized networks can help with agent CNNs focusing on background rather than important objects. By Xingyou Song, , Yilun Du, and

    Prikaži ovu nit
    Poništi
  9. A lot of exciting papers last/this week, probably due to NeurIPS! I summarized a few of them in my weekly newsletter:

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    10. pro 2019.

    Also at our booth, will demo -Agents (). is a reliable, scalable and easy to use RL library. It targets distributed and large-scale RL, and can be easily integrated into your projects and deployed to production.

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet
    3. pro 2019.

    Postdoc positions in theoretical machine learning at Princeton CS Dept. Relevant faculty include Elad Hazan, Ryan Adams, Yoram Singer, and me. Mention in cover letter which faculty you are interested in. Best to apply by Dec 15; latest by Jan 10.

    Poništi
  12. proslijedio/la je Tweet
    4. pro 2019.

    Postdoc positions available at COS and EE Princeton, in the theoretical machine learning group, pls see application details below by .

    Poništi
  13. Poništi
  14. I just realized that everything in my TODO list is inspired by people I follow on Twitter 🧐 1. Swift for TensorFlow (inspired by ) 2. Private AI (inspired by ) 3. Unity ML-Agents (inspired by ) 4. Robot Control (inspired by )

    Poništi
  15. Given raw pixel observations, GAIL discriminators use irrelevant details to discern agent and expert. Let's add a constraint so that the discriminator can't discern agent and expert if there is no meaningful behavior! By and

    Prikaži ovu nit
    Poništi
  16. Using tanh() function immediately after additive noise doesn't work well... let's normalize them first! This replicates the effect and performance of MaxEnt (SAC) on TD3. By C. Wang, Y. Wu, Q. Vuong, and K. Ross

    Prikaži ovu nit
    Poništi
  17. Train off-policy RL, define new macro actions from common action sequences and add them to action space. Repeat! Improves performance of DDPG and SAC on Atari envs: By P. Christodoulou, , , and

    Prikaži ovu nit
    Poništi
  18. Here are three papers I really liked published in arXiv last two weeks:

    Prikaži ovu nit
    Poništi
  19. 3. lis 2019.

    Thoroughly enjoying the RL day! There is a lot to learn

    Poništi
  20. The Appendix of the new Attraction-Repulsion Actor Critic paper is a great example of using this checklist! (Paper by Thang Doan, , Joelle Pineau (who authored the checklist), and )

    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·