Keunwoo Choi

@keunwoochoi

At , after . music, machine learning. My tweets do not represent my employer.

New York, NY
Vrijeme pridruživanja: lipanj 2015.

Tweetovi

Blokirali ste korisnika/cu @keunwoochoi

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

  1. prije 8 sati

    "Looking into tfrecords without Tensorflow" Whaaattttt?

    Poništi
  2. 6. velj

    Try out mirdata! Because writing your own dataloader is boring at best; risky at worst.

    Poništi
  3. 3. velj

    maybe there's a hidden API in tensorflow 0.0.0.1 that can instantiate a general artificial intelligence but no one has figured out how to correctly use it bc like everything else, there's zero documentation hence no one could write a tutorial about it?

    Prikaži ovu nit
    Poništi
  4. 3. velj

    why is tensorflow more complicated then python?????????

    Prikaži ovu nit
    Poništi
  5. 3. velj

    why is tf.records... like this?????????? who designed it? how the hell i am supposed to look into it when i don't exactly know what kinda key it has there?

    Prikaži ovu nit
    Poništi
  6. 29. sij

    Let's make ICASSP affordable! Please share your experience regarding its high registration fee.

    Poništi
  7. 28. sij

    I’d rather let Uber organize a conference. Their review system is fairer; and they’re non-profit :P

    Prikaži ovu nit
    Poništi
  8. 28. sij

    And of course this “FULL MONEY POLICY” remains instead of adopting something we deserve, like, double blind review.

    Prikaži ovu nit
    Poništi
  9. 28. sij

    ICASSP really needs some change.. if not a lot. But I don’t see any reason to be optimistic knowing that this is 2020 and they still didn’t adopt double-blind review system.

    Poništi
  10. proslijedio/la je Tweet
    28. sij

    I'm happy to announce our latest work on self-supervised learning for . PASE+ is based on a multi-task approach useful for recognition. It will be presented at . paper: code: @Mila

    Prikaži ovu nit
    Poništi
  11. 26. sij

    Google's AI Platform is amazing. I mean, only until the training stops for no reason. I mean, it's amazing for about 1-5 hours.

    Poništi
  12. 26. sij
    Poništi
  13. 26. sij
    Poništi
  14. 25. sij

    This icon never looked like "copy to clipboard" to me that I was ranting alone about why Github didn't put one there. Then I was like "so what does this stupid button do instead? (click) (nothing seemed to happen) ..oh wait.. that looks like a clipboard.."

    Poništi
  15. 25. sij

    experienced pretty much the same and i bet others did too. don’t get fooled with the reduced Flops! the speed of depthwise conv can be disappointing!

    Poništi
  16. 22. sij

    especially given the inefficiency of layers like depthwise convolution.

    Poništi
  17. proslijedio/la je Tweet
    16. sij

    torchaudio v0.4: more transforms, datasets, backend support - LibriSpeech and Common Voice loaders - Filters (biquad), batched / jittable transforms (MFCC, gain, dither), more augmentation - interactive speech recognition demo with voice detection

    Poništi
  18. 20. sij

    I’m doing the same for the same reasons.

    Poništi
  19. proslijedio/la je Tweet
    16. sij

    “Deep Learning for Black-Box Modeling of Audio Effects,” newly published Martinez et al in Special Issue of . Paper, code, and dataset:

    Prikaži ovu nit
    Poništi
  20. proslijedio/la je Tweet
    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:

    ·