Martin Ingram

@xenophar

PhD student at University of Melbourne. Studying statistics; background in physics & computer science. Passionate about stats, tennis, and machine learning.

Melbourne, Australia
Vrijeme pridruživanja: veljača 2012.

Tweetovi

Blokirali ste korisnika/cu @xenophar

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

  1. 2. velj

    Only player to take a set off Novak so far is... Jan-Lennard Struff in the first round. Clearly that was the real final

    Poništi
  2. 27. sij

    *slaps roof of neural net* This bad boy can overfit so many datasets

    Poništi
  3. proslijedio/la je Tweet

    Best predatory journal spam invite ever 🤣

    Poništi
  4. 10. pro 2019.
    Poništi
  5. 20. stu 2019.

    Its only p-hacking if it comes from the pihâque region of France. Otherwise it's just sparkling malpractice

    Poništi
  6. 10. stu 2019.

    Anyone have a nice way of composing functions in python? I always end up writing lambdas. E.g. if I only want to get the second element of a function that returns a tuple I do: def second(*args): return other_fun(*args)[1] I feel like there should be a nicer way?

    Poništi
  7. 1. stu 2019.

    Really enjoyed this talk, very thought provoking. MLSS video lectures are really a treasure trove...

    Poništi
  8. 31. lis 2019.

    This looks very interesting. Love the idea of learning the kernel function together with the GP and keen to read in detail

    Poništi
  9. 29. lis 2019.

    Of course there are many many caveats and potential confounders here but I _am_ intrigued by the sharp drop in Zverev's DF rate post-Lendl (one image shows mean only; other shows mean and data)

    Poništi
  10. 23. lis 2019.

    Casella and Berger absolutely torching the Classic ANOVA hypothesis

    Poništi
  11. 14. lis 2019.

    Me: how hard can SPDEs be? *Starts reading SPDE book* ...

    Poništi
  12. 10. lis 2019.

    This is quickly becoming one of my favourite papers, esp section 4. Took me a while to understand but the bound is so simple and easy to implement in modern frameworks! Scalable Variational Gaussian Process Classification Hensman, Matthews, Ghahramani

    Poništi
  13. proslijedio/la je Tweet
    10. lis 2019.
    Poništi
  14. proslijedio/la je Tweet
    8. lis 2019.

    Stich Fix is selling something called pants, a concept from quantum mechanics.

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

    proof by "that seems reasonable to me"

    Poništi
  16. proslijedio/la je Tweet
    1. lis 2019.

    `opt_einsum` v3.1 is out today! A new dynamic programming path has been added that is nearly optimal, but much faster for 5-10 tensors than a true `optimal` implementation.

    Poništi
  17. 28. ruj 2019.

    This is probably one of my prettier accidental plots

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

    Neural reparameterization improves structural optimization! By parameterizing physical design in terms of the (constrained) output of a neural network, we propose stronger and more elegant bridges, skyscrapers, and cantilevers. With shoyer@ samgreydanus@

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

    [New blog post] Tired of getting NaNs and overflow warnings in logistic regression? I compared different approaches and this is what I found out:

    Prikaži ovu nit
    Poništi
  20. 26. ruj 2019.

    I'm interested in learning more about Krylov methods both because they seem useful and because they sound like a forbidden discipline from a science fiction film

    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:

    ·