Ali Rahimi

@alirahimi0

Self-driving AI drones on the blockchain.

Vrijeme pridruživanja: kolovoz 2012.

Tweetovi

Blokirali ste korisnika/cu @alirahimi0

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @alirahimi0

  1. 8. ruj 2019.

    Accept rate seems like a red herring as indicator of conference quality. At a conference I expect to appreciate O(1) papers, not O(r n). If anything, adapting accept rate to number of attendees makes sense to me to make sure everyone gets their O(1).

    Tweet je nedostupan.
    Poništi
  2. 20. kol 2019.

    Here's a clever use of machine learning in astronomy. To predict the location of an asteroid, they fit a complicated parametric model inspired by physics to a few observations of the trajectory. Spoiler: Gauss, 1809.

    Poništi
  3. 20. kol 2019.

    Statistics twitter, is my interpretation correct that Robbins 1951 discovered Stein's phenomenon on ±1 vectors 5 years before Stein discovered it for R^n?

    Poništi
  4. proslijedio/la je Tweet
    29. tra 2019.

    If you are working on empirical phenomena in deep learning, consider submitting to our ICML workshop "Identifying and Understanding Deep Learning Phenomena" (). The deadline is May 5, but relevant work that was already published elsewhere is still welcome!

    Prikaži ovu nit
    Poništi
  5. 6. tra 2019.

    can we just pretend that in some parts of the world, "pleaser consider" is a totally acceptable idiom?

    Prikaži ovu nit
    Poništi
  6. 5. tra 2019.

    my second favorite thing about this talk is that at the last minute, he scrapped the talk he was planning to give, and stayed up late late into the night to write this one, based on conversations he'd had over dinner.

    Prikaži ovu nit
    Poništi
  7. 5. tra 2019.

    A mind-blowing talk by on how we develop deep learning algos, and what it means to "understand" these algos (spoiler: by construction, not understandable, he says)

    Prikaži ovu nit
    Poništi
  8. 5. tra 2019.

    ... and an even better step at getting to know its fabulous organizers, and Ying Xiao.

    Prikaži ovu nit
    Poništi
  9. 5. tra 2019.

    Pleaser consider submitting (or attending) our workshop on Identifying and Understanding Deep Learning Phenomena (). I consider the workshop a step at gaining the kind of understanding I wish we had.

    Prikaži ovu nit
    Poništi
  10. 15. velj 2019.

    protip, maybe: if you really think your technology is so evil that it should be kept in a bottle, maybe don't build it in the first place? otherwise people might think it's just hype& vaporware.

    Poništi
  11. proslijedio/la je Tweet
    25. ruj 2018.

    I envision a field that is neither. A field mindful of its own limitations that approaches other disciplines with respect and curiosity.

    Poništi
  12. proslijedio/la je Tweet
    8. ruj 2018.

    (Thread) Some thoughts about peer review post-NIPS:

    Prikaži ovu nit
    Poništi
  13. 31. kol 2018.

    After squad was declared solved, squad 2 emerged. Maybe similar to ImageNet -> COCO, as pointed out.

    Prikaži ovu nit
    Poništi
  14. 31. kol 2018.
    Prikaži ovu nit
    Poništi
  15. 31. kol 2018.

    some NLP benchmarks:

    Prikaži ovu nit
    Poništi
  16. 31. kol 2018.
    Prikaži ovu nit
    Poništi
  17. 31. kol 2018.

    here are a few more, to tell a more balanced story. Thanks to for help. PASCAL VOC:

    Prikaži ovu nit
    Poništi
  18. 24. kol 2018.

    to clarify, I'm trying to map where DL is in Christensen's technology S-curves:

    Prikaži ovu nit
    Poništi
  19. 24. kol 2018.

    Just plotted this to check if we're plateauing in computer vision performance, and if it's time to invest in a new wave of techniques. Do you find this this convincing? (unsure what human perf is for this)

    Prikaži ovu nit
    Poništi
  20. 7. lip 2018.

    If you have to say one of these things about this paper, 1) we've overfitt to CIFAR10 validation set 2) its validation has been resilient to overfitting, pls pick 2 (the claim is even subtler). I've heard 3 times this week that ML is screwed because of this. We're not screwed.

    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:

    ·