Siarhei S

@sergeiagreat

a very bad mathematician

Minsk, Belarus
Vrijeme pridruživanja: lipanj 2013.

Tweetovi

Blokirali ste korisnika/cu @sergeiagreat

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

  1. 12. sij

    Useless in shell, displays hashes instead of files names with gio list google-drive://<google_email>. , google-gvfs,

    Poništi
  2. 8. sij

    The worst invention ever made is to replace +++ with . What worse is that , , is cheaper than buying your own 10-20K$ DL cluster. WTF?!

    Poništi
  3. proslijedio/la je Tweet
    7. stu 2019.

    arXiv is temporarily unavailable. We are working to get it back up, and apologize for the inconvenience.

    Poništi
  4. 29. lis 2019.

    NameError: free variable 'do_return_2' referenced before assignment in enclosing scope the best tensorflow related error so far .0 .0

    Poništi
  5. 10. lis 2019.
    Poništi
  6. proslijedio/la je Tweet
    8. ruj 2019.

    genie: i will grant you three wishes aladdin: ok genie: but before each wish you must watch an ad aladdin: what genie: to skip the ads you can get Genie Prime aladdin: genie: also i watch you while you sleep aladdin: genie: and tell the other genies your preferences, habits etc

    Poništi
  7. proslijedio/la je Tweet
    10. kol 2019.

    Paged Out! #1 is out! (and it's free to download!) There are 57 articles in 12 categories: Electronics Programming Assembly Reverse Engineering Sec/Hack Retro File Formats Algorithmics SysAdmin Radio Phreaking OS Internals Enjoy! !

    Prikaži ovu nit
    Poništi
  8. 19. kol 2019.

    Quite a moving song! Lukas Graham - 7 Years

    Poništi
  9. 8. kol 2019.

    Nice solution within tensorflow!

    Poništi
  10. 7. kol 2019.

    Var-nal Bayesian Lin. Regr. Setting prior `c` to sqrt(10) leads to an optimal tradeoff between kl_div and minus_elbo. First 3 plots are SGD with c = [1e-1, 1e-2, 1e-3], the rest 3 have c = [1, 10, sqrt(10)]. The very first one is Adam, c=1.

    Poništi
  11. 3. kol 2019.

    A detailed math book on Variational Bayesian Regression. Published in 2012. 1000 Pages.

    Poništi
  12. 20. srp 2019.

    It's not clear why (72c) doesn't have `log` before `u(g^{-1}(x,\Theta);\Theta)`. Given the following equation for differentials.

    Poništi
  13. 4. srp 2019.

    A random projection onto Hamming cube for a NN search. The upper bound on the expected query time is not tight. The following paper obtains tight bounds for any c>1.

    Poništi
  14. 30. lip 2019.

    A tree based representation of a similarity function on product space. Split relies on an entropy based heuristics (see 2.a). Each tree node is a disjoint subset. Nodes are ordered according to a similarity rank.

    Poništi
  15. 20. lip 2019.

    It's not clear what meaning does eps<0.5 have on adversarial accuracy. Since input to models is likely to be RGB with a scale [0, 255]. Authors mention L2-norm to obey power law for \epsilon 1..70, but didn't provide a plot for it. There're no samples for eps > 2 at the plot.

    Poništi
  16. 20. lip 2019.

    Power law of an adversarial error with an increased perturbation error can be related to the ill calibrated NN.

    Poništi
  17. 18. lip 2019.

    A paper on neural networks prunning. It's done at the intialization step. 90% prunned WRN performs same as reference.

    Poništi
  18. 30. tra 2019.

    There's a DL architecture from AspeRa. Which learns aspects among embedded textual reviews into a latent space. Met the author at datafest 2019, Minsk. arxiv [1902.07829]

    Poništi
  19. 3. ožu 2019.
    Poništi
  20. 20. velj 2019.

    Looks like a well written article on features in Gentoo, Cargo, Vcpkg

    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:

    ·