Ari Morcos

@arimorcos

Research Scientist at (FAIR) working to build a science of deep learning. Formerly at . Opinions are my own.

Vrijeme pridruživanja: travanj 2009.

Tweetovi

Blokirali ste korisnika/cu @arimorcos

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

  1. Prikvačeni tweet
    25. stu 2019.

    Excited to share our blog summarizing some of our recent work understanding the boundaries of the lottery ticket hypothesis! Do lottery tickets generalize across datasets? Is the phenomenon present in RL and NLP? Can we begin to explain it theoretically?

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

    Hot take: Mathiness [1] is like an adversarial patch [2] for ML conference reviewers: Mathiness causes a reviewer to classify the paper as "accept" regardless of whether the math is useful/valid and the paper is any good. [3] Fig. 6 has some empirical evidence of this. (refs ⬇️)

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    21. sij
    Prikaži ovu nit
    Poništi
  4. proslijedio/la je Tweet

    Idle reminder that ‘I don’t know’, ‘I don’t have an opinion’ or ‘that’s not my field’ are all perfectly good answers

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

    Some people in ML share the illusion that models expressed symbolically will necessarily/magically generalise better compared to, for example, parametric model families fit on the same data. This belief seems to come from a naive understanding of mathematics 1/5

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

    Ever wondered what happens when you freeze all the weights in a neural network and only train batch normalization? Me too! Turns out you can get 80%+ accuracy on CIFAR-10 by doing so. Check out our poster and oral in the SEDL workshop in West 121. With David Schwab and

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

    It is commonly said that models like BERT or GPT-2 don't really 'understand', but what does it actually mean to understand language? We try to answer this via a roadmap for human-like understanding of language in machines.

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

    Do lottery ticket initializations generalize, or are they overfit to the precise conditions used to generate them? If you're at , come see our poster #170 happening right now! Paper: Blog:

    Poništi
  10. proslijedio/la je Tweet

    We’re presenting our paper on understanding the generalization of the “lottery ticket” phenomenon at . Read about it here: . Visit our poster today at 10:45 AM -- 12:45 PM @ East Exhibition Hall B + C #170.

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

    Presenting "A Benchmark for Interpretability Methods in Deep Neural Networks" at poster #159 tomorrow (Dec 10th). Poster > Slides > Code > Paper > See you there! :)

    Poništi
  12. 9. pro 2019.

    Andrew is an awesome scientist, and I'm sure he'll be a thoughtful and supportive mentor. If you're looking for opportunities in DL theory/understanding/neuro, this is a great opportunity!

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

    Excited to present our work on Unsupervised Learning of Object Dynamics from Videos at . Stop by East Exhibition Hall B + C poster 149 Tue 10:45am! with Chen Sun, , , Kevin Murphy,

    Poništi
  14. proslijedio/la je Tweet

    Facebook AI is getting ready to welcome all of you at ! Swing by booth 509 to learn more about cutting-edge research we’re presenting this year. Read more:

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

    How does transfer learning for medical imaging affect performance, representations and convergence? Check out the blogpost below and our paper for some of the surprising conclusions, new approaches and open questions!

    Poništi
  16. 6. pro 2019.

    Excited to be in Vancouver next week for ! Reach out if you want to chat!

    Poništi
  17. proslijedio/la je Tweet

    To advance research on AI that can understand others’ points of view and collaborate effectively, Facebook AI has developed a bot that sets a new state of the art in Hanabi, a card game in which all players work together.

    Poništi
  18. 4. pro 2019.
    Prikaži ovu nit
    Poništi
  19. 4. pro 2019.
    Prikaži ovu nit
    Poništi
  20. proslijedio/la je Tweet
    1. pro 2019.

    GM is the root cause of the purported war between connectionists and symbolists He created it to make himself relevant. But it does not actually exist. Clamoring that DL should be hybridized with symbol manipulation is useless. Showing how to do it would be useful.

    Poništi
  21. proslijedio/la je Tweet
    29. stu 2019.

    Find a back door solution? I thought that's what they have to *stop* doing 😄

    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:

    ·