Yigit  

@yigitdemirag

PhD student at INI of ETH Zurich and UZH | neuromorphic computing | previously , , , ,

Zurich
Vrijeme pridruživanja: srpanj 2010.

Tweetovi

Blokirali ste korisnika/cu @yigitdemirag

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

  1. Prikvačeni tweet
    4. pro 2019.

    Unpopular opinion: Backpropagation is neither an elegant nor an interesting algorithm for neural networks. But it works fantastically. Because it's a simple chain rule.

    Prikaži ovu nit
    Poništi
  2. 15. sij

    Postdoc Me Now - A "Don't Stop Me Now" Science Parody (credits to 🤣)

    Poništi
  3. 10. sij

    When your name sounds a bit complicated for barista. – mjesto: Zürich Airport (ZRH)

    Poništi
  4. 9. sij

    Two colors in one minute.

    Poništi
  5. proslijedio/la je Tweet
    2. sij

    This is just huge. And perhaps just the tip of the iceberg; individual neurons may do lots we haven't credited them with doing. & even a single human-specific innovation like this could be deeply illuminating. & it may really reshape our thoughts on "neurally-inpired" networks.

    Prikaži ovu nit
    Poništi
  6. 31. pro 2019.

    Mutlu yıllar Türkiye! 🎉

    Poništi
  7. 30. pro 2019.

    Morning work at 7:30 train to Lausanne.

    Poništi
  8. 24. pro 2019.

    (3/3) Ben de amfide, jüri üyesi hocamın teknik sorusuna, "Maalesef hocam henüz Fransızca konuşamıyorum." diyerek cevap verdim 😭. Ardından büyük sessizlik...

    Prikaži ovu nit
    Poništi
  9. 24. pro 2019.

    (2/3) O sırada her ne sebeptense hocanın sorusunu "Fransızca konuşabiliyor musun?" olarak anladım. Bu soruyu garipsesem de çok mantıksız gelmedi çünkü doktora için İsviçre'nin Fransızca konuşulan Lausanne bölgesine gitmeyi düşündüğümü önceden söylemiştim.

    Prikaži ovu nit
    Poništi
  10. 24. pro 2019.

    (1/3) Master tez savunmam sırasında jüri üyelerinden biri Wiedemann-Franz Law' un genelde metallerdeki termal iletkenligi modellemek için kullanıldığını, benim neden GST modeli için kullandığımı sordu.

    Prikaži ovu nit
    Poništi
  11. 19. pro 2019.
    Poništi
  12. 18. pro 2019.

    First I thought LIF and STDP mechanisms are emerged through evolution on RNNs and super excited but appereantly this is not the case.. They basically trained ENUs first to match LIF and STDP then show that it works 🤷🏼‍♂️

    Poništi
  13. proslijedio/la je Tweet

    Illusion of The Year 2019 Winner It's based on a simple Lissajous curve, but uses clever shading/highlighting to create a 'double-axis' illusion

    Prikaži ovu nit
    Poništi
  14. 15. pro 2019.
    Poništi
  15. proslijedio/la je Tweet
    13. pro 2019.

    “Leaders in artificial intelligence warn that progress is slowing, big challenges remain, and simply throwing more computers at a problem isn't sustainable.”

    Prikaži ovu nit
    Poništi
  16. 14. pro 2019.

    Watched Ad Astra last night. Really liked the visuals but scenario was too weak and decorated with pseudo-scientific details. 🤷🏼‍♂️

    Poništi
  17. 4. pro 2019.

    I would argue, even bet 100$, that in the next 10 years, we will invent a new learning rule that will have similar accuracy to BP on the scale, but biologically plausible, more efficient and nothing like BP.

    Prikaži ovu nit
    Poništi
  18. 4. pro 2019.

    How iteratively we have engineered the computational fabric in the last 50 years is nothing but an art.

    Poništi
  19. 30. stu 2019.

    I stumbled upon the Diagon Alley this evening.

    Poništi
  20. 23. stu 2019.
    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:

    ·