Zizhao Zhang

@ZizhaoZhang

Research@Google

Sunnyvale, CA
Vrijeme pridruživanja: travanj 2012.

Tweetovi

Blokirali ste korisnika/cu @ZizhaoZhang

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

  1. prije 7 sati

    FixMatch ImageNet! A single model training yields 71.5% top-1 accuracy.

    Poništi
  2. proslijedio/la je Tweet
    prije 8 sati
    Poništi
  3. proslijedio/la je Tweet
    31. sij

    A well done video explanation of FixMatch, thanks !

    Poništi
  4. 22. sij

    Our new method on semi-supervised learning.

    Poništi
  5. 12. sij
    Poništi
  6. proslijedio/la je Tweet
    5. sij

    Progress isn’t always easy to see. These charts show some of the ways that the world is getting better.

    Poništi
  7. proslijedio/la je Tweet

    Nice slides and pointers! Sim is not real, but a "widened enough" sim (using enough augmentation) contains real as an element, somewhat.

    Poništi
  8. proslijedio/la je Tweet
    10. stu 2019.

    😊Self-supervised learning opens up a huge opportunity for better utilizing unlabelled data while learning in a supervised learning manner. My latest post covers many interesting ideas of self-supervised learning tasks on images, videos & control problems:

    Poništi
  9. proslijedio/la je Tweet
    28. lis 2019.

    Thanks Augustus! Enjoy working with you on GANs again :)

    Poništi
  10. proslijedio/la je Tweet
    28. lis 2019.

    has new research (w/ , and me) that sets a new state-of-the art in conditional image synthesis by using consistency regularization on GANs: . Thread follows:

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet

    We're sharing a new approach for teaching AI how to interact with objects by showing it videos of everyday behavior. This research uses weak supervision and teaches systems to understand interaction hotspots.

    Poništi
  12. proslijedio/la je Tweet
    10. ruj 2019.

    It's hard to scale meta-learning to long inner optimizations. We introduce iMAML, which meta-learns *without* differentiating through the inner optimization path using implicit differentiation. to appear w/

    Poništi
  13. proslijedio/la je Tweet

    The first-ever fastMRI challenge is here! Based on a research project by Facebook AI and NYU Langone Health, this challenge aims to produce diagnostic-quality images in less time. Winning teams will be invited to present at a NeurIPS 2019 workshop.

    Poništi
  14. proslijedio/la je Tweet
    19. srp 2019.

    Advances in show great promise for assisting in the work of healthcare professionals. Check out SMILY, an ML-based similar image search tool for anatomic pathology, along with research into real-world usability.

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

    Our new paper: Large Memory Layers with Product Keys We created a key-value memory layer that can increase model capacity for a negligible computational cost. A 12-layer transformer with a memory outperforms a 24-layer transformer, and is 2x faster! 1/2

    Prikaži ovu nit
    Poništi
  16. proslijedio/la je Tweet
    26. lip 2019.

    I would like to announce that are releasing image augmentation library v0.3.0. To simplify the reproducible we added load / save to python dictionary, json, and yaml config files.

    Poništi
  17. proslijedio/la je Tweet
    18. lip 2019.

    Together w/ colleagues at & we propose dynamically optimized acquisition for faster, more accessible . in

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    18. lip 2019.

    Faster MRI through active acquisition. FAIR+NYU Langone Medical Center.

    Poništi
  19. 18. lip 2019.

    Our paper for active acquisition in undersampled MRI reconstruction, a part of the project, is live. – mjesto: Long Beach Convention & Entertainment Center

    Poništi
  20. proslijedio/la je Tweet
    18. lip 2019.

    and Matt presenting our work on active acquisition for MRI reconstruction now at . Stop by poster 207!

    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:

    ·