Adam Kosiorek

@arkosiorek

PhD candidate at Oxford and a Research Scientist at DeepMind; I'm trying to understand what intelligence is and to implement it.

London, England
Vrijeme pridruživanja: prosinac 2014.

Tweetovi

Blokirali ste korisnika/cu @arkosiorek

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

  1. proslijedio/la je Tweet
    prije 4 sata

    camera-ready version of GENESIS is now online: We would like to sincerely thank the reviewers for their thoughtful comments and useful suggestions :)

    Poništi
  2. proslijedio/la je Tweet
    30. sij

    We are planning to hold events, big and small, for in Ethiopia, in Austria, in Boston, in Vancouver, and more! Apply to be an organizer if you'd like to be more involved.

    Poništi
  3. 15. sij

    Congrats to ! I've learned a lot from this guy and would totally recommend working with him for all things statistical ML!

    Poništi
  4. proslijedio/la je Tweet
    12. sij

    In prior work (Doe et al, 2019) has considered the problem of parrot walking, however, the proposed method had severe limitations. The approach presented in this paper is novel and versatile. To our knowledge it is the first work considering multiple parrots simultaneously.

    Poništi
  5. 8. sij

    I'm excited to announce that I've just joined DeepMind in London as a Research Scientist in 's team. I'm planning to continue working on unsupervised representation learning -- let's see how far we can scale object-centric representations!

    Poništi
  6. 30. pro 2019.

    Now, I'm about to graduate after about three years at Oxford, and I can keep doing whatever research I like, and wherever I want. What's the point of dragging the PhD out into 6+ years total? 5/5

    Prikaži ovu nit
    Poništi
  7. 30. pro 2019.

    Is it really this much worse to get a PhD at a slightly worse school, and to prove your worth with your research work? I had no prior ML work when I applied for PhD, and I got rejected from Stanford and from Toronto (which isn't even considered "top", I guess). 4/5

    Prikaži ovu nit
    Poništi
  8. 30. pro 2019.

    Sure, these schools are prestigious, but, as US institutions, they often require 4-6 years to finish the PhD. You also need to account for the time needed to get these additional publications. 3/5

    Prikaži ovu nit
    Poništi
  9. 30. pro 2019.

    Getting a PhD is only the first stage of a scientific career, and its purpose is to learn how to do research. If you need to know how to do research to even start, then I'm not sure what the point is. 2/5

    Prikaži ovu nit
    Poništi
  10. 30. pro 2019.

    Apparently, you need 1-2 publications at top-tier ML conferences to get into a top school for a PhD: I wonder if getting into top schools is still worth it, then? 1/5

    Prikaži ovu nit
    Poništi
  11. 28. pro 2019.

    Is worth it over for an undergrad in CS? Asking for my little bro :)

    Poništi
  12. 21. pro 2019.

    GENESIS just got accepted to . Great job and the team; see you all in Ethiopia! :)

    Poništi
  13. 17. pro 2019.

    what's the difference between density estimation and generative modelling?

    Poništi
  14. 17. pro 2019.
    Prikaži ovu nit
    Poništi
  15. 17. pro 2019.

    The idea is that you can predict a set from any conditioning by: 1. Encoding the conditioning. 2. Encoding a randomly initialised set of some maximum size. 3. Running gradient descent on the set to match its encoding to the encoding of the conditioning.

    Prikaži ovu nit
    Poništi
  16. 16. pro 2019.

    A nice summary of : . Interestingly, it's almost completely orthogonal to my experience...

    Poništi
  17. 16. pro 2019.

    This paper had the biggest subjective "wow" factor for me at : "Deep Set Prediction Networks" by Yan Zhang, Jonathon Hare, Adam Prügel-Bennett . The idea is very simple, and yet I find it mind-blowing.

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

    Come to our poster tomorrow (Saturday) on relational reasoning in multi-object tracking! ✌️ 9:45 - 10:30, Sets & Partitions WS, West Level 2, #215-216 See a video of results on real-world data: Work with

    Poništi
  19. 6. pro 2019.

    Going to ? Come see us on Thursday in the East Exhibition Hall B+C at 5pm, where we're presenting Stacked Capsule Autoencoders (poster #36)!

    Poništi
  20. 6. pro 2019.

    Ever wondered how to do meta-learning directly in function space? Here's a treat from with soma SOTA results on mini/tiredImageNet!

    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:

    ·