Aravind

@Aravind7694

Interested in building simple ML stuff that works efficiently at scale, PhD , Interned at and .

Berkeley, CA
people.eecs.berkeley.edu/~aravind
Vrijeme pridruživanja: srpanj 2016.
Rođen/a 07. lipnja
4 fotografije ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @Aravind7694

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @Aravind7694

  1. 31. sij

    I read this advice 1.5 years ago and have stuck with it since then. It's not mutually exclusive of course. But having clear goals / OKRs (even for new ideas) allows for a sense of direction and purpose and also avoid writing papers that don't matter in the long run.

    An Opinionated Guide to ML Research: “To make breakthroughs with idea-driven research, you need to develop an exceptionally deep understanding of your subject, and a perspective that diverges from the rest of the community—some can do it, but it’s hard.”
    Prikaži ovu nit
    1 reply 15 proslijeđenih tweetova 121 korisnik označava da mu se sviđa
    Poništi
  2. 29. sij

    Robotic Learning done right with full stack execution :-) ...

    Couldn’t be more excited about what we are building! Yet, even more excited about team, all brilliant and exceptional at working across team boundaries, engineering, research, business, we are all having fun and getting it done together!
    1 proslijeđeni tweet 25 korisnika označava da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    26. sij

    Teaching Deep Unsupervised Learning (2nd edition) at this semester. You can follow along here: Instructor Team: , , , Wilson Yan, Alex Li, YouTube, PDF, and Google Slides for ease of re-use

    Korisnici , i
    121 proslijeđeni tweet 413 korisnika označava da im se sviđa
    Poništi
  4. proslijedio/la je Tweet
    19. sij

    Crew Dragon separating from Falcon 9 during today’s test, which verified the spacecraft’s ability to carry astronauts to safety in the unlikely event of an emergency on ascent

    1.956 replies 24.778 proslijeđenih tweetova 147.856 korisnika označava da im se sviđa
    Poništi
  5. proslijedio/la je Tweet
    19. pro 2019.

    We’ve just had the best decade in human history. Seriously

    51 reply 475 proslijeđenih tweetova 1.224 korisnika označavaju da im se sviđa
    Poništi
  6. proslijedio/la je Tweet
    8. sij

    Just ordered a couple 1TB thumb drives for $30 each. I still find myself awestruck by tech progress -- I desperately wanted a 10MB hard drive for my IIGS as a teen, but didn't have the $399. One million times cheaper per byte, one thousand times faster, and 100 times smaller now.

    190 replies 663 proslijeđena tweeta 5.890 korisnika označava da im se sviđa
    Poništi
  7. 7. sij

    "When running on the test set, examples were randomly perturbed using the same augmentations as during training; the final predictions were the average of 500 runs." - Test-time augmentations on steroids. [s/500/1500; ensemble of three models -1 obj detector with ~15k ROI labels

    Check out new research published in that evaluates the accuracy of a model developed to detect breast cancer in de-identified screening mammograms. Read the full paper: Read the Keyword article:
    1 reply 12 korisnika označava da im se sviđa
    Poništi
  8. proslijedio/la je Tweet
    20. pro 2019.

    In case you have missed it, here is my end-of-year AI/ML recap with over 50 pointers to publications:

    19 proslijeđenih tweetova 76 korisnika označava da im se sviđa
    Poništi
  9. proslijedio/la je Tweet
    20. pro 2019.

    I often hear people say that X is broken. Congress is broken. Academia is broken. Physics is broken. Deep learning is broken. People like to stand in crowds and point fingers at the thing that is broken. I prefer the quiet folks who roll up their sleeves and get to work, fixing.

    139 replies 741 proslijeđeni tweet 4.147 korisnika označava da im se sviđa
    Poništi
  10. 19. pro 2019.

    Cool Medium blogpost explaining key aspects and results of the self-supervised pre-training pipeline CPC-v2:

    6 proslijeđenih tweetova 23 korisnika označavaju da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    16. pro 2019.

    Lost amid NeuroIPS: jax now has experimental cloud TPU support!

    34 proslijeđena tweeta 149 korisnika označava da im se sviđa
    Poništi
  12. proslijedio/la je Tweet
    9. pro 2019.

    There are now 3 papers that successfully use Self-Supervised Learning for visual feature learning: MoCo: PIRL: And this, below. All three use some form of Siamese net.

    Unsupervised pre-training now outperforms supervised learning on ImageNet for any data regime (see figure) and also for transfer learning to Pascal VOC object detection
    355 proslijeđenih tweetova 1.094 korisnika označavaju da im se sviđa
    Poništi
  13. proslijedio/la je Tweet
    9. pro 2019.

    Exciting updated results for self-supervised representation learning on ImageNet: - 71.5% top-1 with a *linear* classifier - 77.9% top-5 with only *1%* of the labels - 76.6 mAP when transferred to PASCAL VOC-07 (better than *fully-supervised's* 74.7 mAP)

    1 reply 12 proslijeđenih tweetova 74 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    8. pro 2019.

    Some exciting *new* results in self-supervised learning on ImageNet: 71.5 % top-1 with a linear classifier, 5x data-efficiency from pre-training (76% top-1 with 80% fewer samples per class on ImageNet), 76.6 mAP on PASCAL VOC-07 (> supervised's 74.7)

    55 proslijeđenih tweetova 258 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    9. pro 2019.

    Beating previous state of the art in self-supervised learning for ImageNet by almost 3% absolute with less parameters (71.5% vs 68.6% top1). Extensive results for data-efficient learning on both ImageNet and Pascal VOC in the updated

    , , i još njih 2
    5 proslijeđenih tweetova 14 korisnika označava da im se sviđa
    Poništi
  16. proslijedio/la je Tweet
    9. pro 2019.

    Unsupervised pre-training now outperforms supervised learning on ImageNet for any data regime (see figure) and also for transfer learning to Pascal VOC object detection

    , , i još njih 2
    218 proslijeđenih tweetova 778 korisnika označava da im se sviđa
    Poništi
  17. 8. pro 2019.

    With , , , , and .

    1 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  18. 8. pro 2019.

    Some exciting *new* results in self-supervised learning on ImageNet: 71.5 % top-1 with a linear classifier, 5x data-efficiency from pre-training (76% top-1 with 80% fewer samples per class on ImageNet), 76.6 mAP on PASCAL VOC-07 (> supervised's 74.7)

    55 proslijeđenih tweetova 258 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. proslijedio/la je Tweet
    11. stu 2019.

    Any friends in mgmt/exec/leadership roles you wish knew more about AI + benefit from the push of a formal class (they can take from home :) ? I just finished my recordings for 's Artificial Intelligence: Business Strategies and Applications

    21 proslijeđeni tweet 114 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  20. 3. lis 2019.

    This is a cool paper demystifying the myth that we need model-based methods for data-efficiency. Turns out Rainbow with correct hyper-parameters is as good-

    19 proslijeđenih tweetova 142 korisnika označavaju da im se sviđa
    Poništi

@Aravind7694 još nije objavio nijedan Tweet.

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:

    ·