Louis Kirsch

@LouisKirschAI

AI researcher (PhD) at IDSIA with . Working on RL agents that meta-learn their own RL algorithm. Formerly , .

Vrijeme pridruživanja: studeni 2011.

Tweetovi

Blokirali ste korisnika/cu @LouisKirschAI

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

  1. Prikvačeni tweet
    31. lis 2019.

    Meta Reinforcement Learning is good at adaptation to very similar environments. But can we meta-learn general RL algorithms? Our new approach MetaGenRL is able to. With and Paper: Blog:

    Poništi
  2. prije 2 sata

    The deadline for our workshop Beyond 'Tabula Rasa' in RL (BeTR-RL) is getting closer. We are looking forward to all contributions! February 10th 2020, 11:59 PM anywhere on Earth

    Prikaži ovu nit
    Poništi
  3. 13. sij

    We are excited to announce our workshop 'Beyond "Tabula Rasa" in RL (BeTR-RL): Agents that remember, adapt, and generalize' w/ Ignasi Clavera, Kate Rakelly, , , CfP & more infos:

    Prikaži ovu nit
    Poništi
  4. 30. pro 2019.

    I couldn't agree more “One must keep in mind that the hard part is not getting the Grand Idea, but demonstrating that it can work, turning it into a practical algorithm or method, and perhaps even beat the state of the art on some practical problem.“

    Poništi
  5. 14. pro 2019.

    Location changed to Ballroom D

    Prikaži ovu nit
    Poništi
  6. 14. pro 2019.

    I'm organizing a small meetup with to chat about meta learning and automatically generating environments. Please come and join us! I will also present my work on meta learning new RL algorithms that generalize to vastly different environments.

    Prikaži ovu nit
    Poništi
  7. 12. pro 2019.

    I'm really looking forward to this talk! It fits well with my research on meta learning general RL algorithms

    Poništi
  8. 11. pro 2019.

    This paper is an absolute highlight! Forget batch norm, forget residual connections, a low gradient quotient initialization is all you need.

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

    Blaise Aguera ()’s keynote talk at suggests that the machine learning research community employ ideas from artificial life (evolution, cellular automata, Hebbian learning etc.) to address fundamental limitations of deep learning.

    Prikaži ovu nit
    Poništi
  10. 4. lis 2019.

    A great overview if you'd like to know more about Juergen's work in the early 90s!

    Poništi
  11. 26. ruj 2019.
    Poništi
  12. proslijedio/la je Tweet

    Most of us know Schmidhuber through papers, anecdotes, and memes. I thought this conversation between him and was fascinating and worth listening from start to finish:

    Prikaži ovu nit
    Poništi
  13. 13. ruj 2019.

    Sapiens is indeed a very good read! Society is based on stories.

    Poništi
  14. 12. ruj 2019.

    My supervisor Jürgen Schmidhuber is now on Twitter!

    Poništi
  15. proslijedio/la je Tweet
    1. srp 2019.

    Animal-AI Olympics v1.0 released! Launch info with prizes and category descriptions at Submissions slightly delayed, opening July 8th @ EvalAI. Thanks to our sponsors and collaborators! Including:

    Poništi
  16. proslijedio/la je Tweet
    5. kol 2019.
    Poništi
  17. proslijedio/la je Tweet
    21. svi 2019.

    Awesome map of Reinforcement Learning concepts - A must read! "All in all, it is easy to get lost in the large number of publications and subfields of research." You said it! This kind of bird's-eye view is sorely missing in DRL.

    Poništi
  18. proslijedio/la je Tweet
    30. lis 2018.

    MAX: a new algorithm from our team for actively taking decisions that lead to new information about an environment. The goal of this research is to substantially speed up exploration for Model-Based Reinforcement Learning. Model-Based Active Exploration

    Prikaži ovu nit
    Poništi
  19. 13. tra 2019.

    Incredible what PPO with massive batch sizes and long training can do. It was very interesting to watch the 2-0 win as an AI researcher.

    Poništi
  20. proslijedio/la je Tweet
    10. tra 2019.

    Schmidhuber's Low-complexity Art (1994)

    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:

    ·