HenryE

@EldinhoC

Doing a PhD in modelling data-to-text systems using deep learning

Vrijeme pridruživanja: svibanj 2013.

Tweetovi

Blokirali ste korisnika/cu @EldinhoC

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

  1. proslijedio/la je Tweet
    5. stu 2019.

    The Erasure online hackathon is live! Register at and join the Numerai team online this Thursday morning for a live hackathon workshop ⛓️

    Poništi
  2. proslijedio/la je Tweet
    31. lis 2019.

    Why do people use Medium? I don't get why people are hiding their content behind someone else's paywall. What's amazing is how many tech startups have their official blog on Medium. So I'm supposed to pay to read your company's blog??

    Poništi
  3. 30. lis 2019.

    The changes made to comments a while ago were bad enough. Namely; not being able to highlight text within a comment without clicking it & opening a new page, forcing links in comments to be opened in the same tab (even when using cmd+click), and making comments separate to blog

    Prikaži ovu nit
    Poništi
  4. 30. lis 2019.

    How on earth is getting away with paywalling so much user generated content related to ML? This is like if stackoverflow started paywalling

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet

    sharing publicly for the first time, an in depth look at the backtest of Numerai's meta model.

    Poništi
  6. 29. ruj 2019.

    What's interesting with these Summon videos is that they seem more like failures of unspoken communication between drivers and theory of mind than actual technical driving errors

    Poništi
  7. proslijedio/la je Tweet
    27. ruj 2019.

    Soday 1 with V10 Smart Summon was working beautifully. But someone didn’t notice my M3 and made a front bumper damage. We will claim our insurances but who’s fault do you guys think it’ll be ? Should I present this videos ?

    Prikaži ovu nit
    Poništi
  8. proslijedio/la je Tweet

    100,000 stock “photos” of non-existent people hallucinated by an AI, copyright-free with attribution.

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

    anyone who has been interested in my games but fancies lower complexity, P1 SELECT might be your thing: it's in similar conceptual space to 868-HACK, Cinco Paus, Imbroglio but with a lot less content.

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

    Hide yourself - arkit testing w

    Poništi
  11. proslijedio/la je Tweet
    21. lip 2019.

    A few of the sculptures created using my Form network on display at the show. The bronze, "Human Allocation of Space", 50 x 68 x 25 cm, is quite striking in real life (true 3D is amazing). Details:

    Poništi
  12. proslijedio/la je Tweet
    26. tra 2019.

    Obligatory animated chart of the astonishing disparity in housing supply between Tokyo and London-NYC-Paris.

    Poništi
  13. proslijedio/la je Tweet
    26. tra 2019.

    Excited to show our recent work on DeepRL for Cassie. Sim2real for a large agile biped. Walking in all directions, various styles, highly robust. An iterative RL design process. An excellent collaboration between UBC / Oregon State / Agility Robotics.

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

    This part of 's talk where adding some shadow lines transforms the output of a pix2pix model is astonishing! The whole video here:

    Poništi
  15. proslijedio/la je Tweet

    I also made the 100 largest islands of the world in minimalist grey. Original relief shading method by . Again, please support us on Patreon! :)

    Poništi
  16. proslijedio/la je Tweet

    Found a hideaway in the wastelands of .

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    19. stu 2018.

    i think this is the first time a pretrained neural net has captured so many creative peoples’ attention. BigGAN has enough capacity/expressive power that people are just exploring the latent space without worrying about making it “their own” (retraining would be very difficult)

    Poništi
  18. 20. kol 2018.

    Really amazing how well NNs can learn to generate high quality output from semi-detailed intermediary representations

    Poništi
  19. 16. kol 2018.

    oh, I decided to go with Babelfy + Babelnet since I'm looking to get hypernyms for relevant words in a sentence and their API seems like a reasonable enough way to do it

    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:

    ·