Jason Ge

@Jian413

Cofounder at Snark AI

San Jose, CA
Vrijeme pridruživanja: travanj 2014.

Tweetovi

Blokirali ste korisnika/cu @Jian413

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @Jian413

  1. proslijedio/la je Tweet
    26. sij

    封鎖と外出自粛により車も人も殆どいない中国武漢市内の様子 新型コロナウイルスの流行で封鎖された武漢市内では、これはチャンスだ!とばかりにカメラやドローンを持って普段と違う街を撮影して廻る人が結構いるみたい

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    12. sij

    been watching this all morning...great work by Yuanming Hu a perfect example showing that a differentiable and accurate simulator allows one to bypass reinforcement learning and use gradient ascent to do planning

    Poništi
  3. proslijedio/la je Tweet
    1. sij

    ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 100%

    Poništi
  4. proslijedio/la je Tweet
    1. sij

    ░░░░░░░░░░░░░░░ 0%

    Poništi
  5. proslijedio/la je Tweet
    29. pro 2019.

    Time for some retrospective! I shared some 25 papers that I particularly enjoyed in the last decade. I would love for you to share some papers that are missing in this list (there are many!!), either here or in the comments on the blog.

    Poništi
  6. proslijedio/la je Tweet
    21. pro 2019.

    This is something in my mind for a long time: "Reinforcement Learning as Posterior Inference" is cool but it is not dealing with the posterior really needed for exploration (which captures epistemic uncertainty). Really great work!

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

    Our paper on the Global Optimality and Convergence Rates of Neural Policy Gradient Methods just got in ICLR 2020 (, )! See you in Addis Ababa, Ethiopia~

    Poništi
  8. proslijedio/la je Tweet

    👋 Hi Friends! I've been hacking on a fun little project. Check out ✨ A directory of creators and all of their best tweets, blogposts, interviews, and podcasts — in one place! 👉 Why? 2-min thread below 👇

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

    How to incorporate exploration into policy optimization? Our recent paper shows that a bonus added to the critic (Q-function) suffices. Check out Optimistic Proximal Policy Optimization (OPPO), which provably explores! See you at the OptRL workshop!

    Poništi
  10. proslijedio/la je Tweet
    Poništi
  11. proslijedio/la je Tweet
    9. pro 2019.

    Announcing . Automatic ML for images, text, tabular data, network architecture search, HPO. Apache open source license. Works on . And , too. With tutorials. Just in time for .

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

    Congratulations to our colleague Lin Xiao for the test of time award!!! Online convex optimization and mirror descent for the win!! (As always? :-).)

    Poništi
  13. proslijedio/la je Tweet

    The first full paper on after 3 years of development. It describes our goals, design principles, technical details uptil v0.4 Catch the poster at Authored by , et. al.

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

    Normalizing Flows let you build up complex, yet still easy to work with probability distributions. Want to learn more? Check out this video I made covering the basics of this growing class of techniques with an example application in generative modeling.

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

    Remember that video about how block collisions can compute the digits of pi? A friend, Adam Brown, just showed that the math underlying this is actually identical to the math behind a very famous quantum search algorithm (Grover's): Genuinely crazy!

    Poništi
  16. proslijedio/la je Tweet
    5. pro 2019.

    This is the first time I use twitter to advertise my research 😅: MT-DNN-SMART (Microsoft D365 AI & Microsoft Research AI & GATECH) achieves new SOTA results in 5 of 9 GLUE benchmark tasks and an overall GLUE task performance 89.9.

    , , i još njih 4
    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    5. pro 2019.

    Meet the 'double descent' phenomenon. After we figure it out we should probably rewrite the book chapter on bias-variance tradeoff.

    Poništi
  18. proslijedio/la je Tweet
    31. lis 2019.
    Poništi
  19. proslijedio/la je Tweet
    18. stu 2019.

    Few people seem to realize that master is easily built for 's ROCm. For this reason and my own reference I wrote a short how-to at If there is enough interest, I could probably also put up binaries for selected Python versions.

    Prikaži ovu nit
    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:

    ·