Brian Cheung

@thisismyhat

This is my hat, there are many like it, but this one is mine.・PhD Student at

Vrijeme pridruživanja: lipanj 2015.
19 fotografija ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @thisismyhat

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

  1. Prikvačeni tweet
    15. velj 2019.

    It turns out you can fit a lot more than one model in one set of parameters...and train them independently. Work with A. Terekhov, Y. Chen, , and B. Olshausen

    5 replies 76 proslijeđenih tweetova 295 korisnika označava da im se sviđa
    Poništi
  2. proslijedio/la je Tweet
    27. pro 2019.

    I'm curious. If you're in machine learning, do you know what the Levels of Analysis are? If so, do you think they are important or relevant for research today in machine learning?

    18 replies 41 proslijeđeni tweet 32 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  3. 23. pro 2019.

    Code to reproduce our NeurIPS 2019 "Superposition of many models into one" () now available:

    1 reply 11 proslijeđenih tweetova 69 korisnika označava da im se sviđa
    Poništi
  4. 22. pro 2019.

    There's a long list of people I know that have left the field/quit their Ph.D. after trying to reproduce 'mindshare'.

    ML is a community of makers. Instead of reproducibility we should optimize for mindshare.
    Prikaži ovu nit
    1 reply 2 proslijeđena tweeta 35 korisnika označava da im se sviđa
    Poništi
  5. 7. pro 2019.

    One sentence summary of all 1427 NeurIPS 2019 papers.

    9 proslijeđenih tweetova 25 korisnika označava da im se sviđa
    Poništi
  6. 18. stu 2019.

    aka physical neurons?

    1 reply 8 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  7. 18. stu 2019.

    aka a spike?

    5 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  8. 18. stu 2019.

    The new Cerebras chip is the most accidentally neuromorphic chip ever.

    1 reply 21 proslijeđeni tweet 67 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    15. lis 2019.

    Our weekly CompNeuro journal club is back up and running for the quarter. Today, we discussed Cheung et al’s preprint "Superposition of many models into one" from and . Led by AJ Kruse and Satpreet Singh.

    11 proslijeđenih tweetova 17 korisnika označava da im se sviđa
    Poništi
  10. 11. tra 2019.

    Glad to see our work making it to modern times . 👏 to for the clean Lasagne/Theano example back in 2015.

    I implemented a 2.0 version of the autoencoder from the paper "Discovering Hidden Factors of Variation in Deep Networks" by Brian Cheung() et al. 2014 😁
    1 reply 11 korisnika označava da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    15. velj 2019.

    Is it to possible to store multiple neural network models within a *single* set of parameters? Yes, No .. maybe? checkout --

    1 reply 24 proslijeđena tweeta 140 korisnika označava da im se sviđa
    Poništi
  12. 31. sij 2019.

    Finally, a definition of disentangling that isn't purely statistical... (Higgins, Amos, , et. al.)

    1 reply 4 proslijeđena tweeta 10 korisnika označava da im se sviđa
    Poništi
  13. 7. sij 2019.

    The revolution will not be optimized.

    If it is possible for automated search methods to invent new concepts such as DropOut, these search methods will probably not be gradient-based. So if such innovative search methods were invented, they probably won’t be used to come up with better gradient-based learning methods.
    5 proslijeđenih tweetova 35 korisnika označava da im se sviđa
    Poništi
  14. 19. pro 2018.

    Amazing how far adding a little bit of structure goes: "from the previous state-of-the-art of 22% to an unprecedented 74%"

    5 proslijeđenih tweetova 16 korisnika označava da im se sviđa
    Poništi
  15. 21. stu 2018.

    When authors fail to do a proper literature review:

    8 korisnika označava da im se sviđa
    Poništi
  16. 17. stu 2018.

    ICLR 2020 in Africa!

    1 proslijeđeni tweet 7 korisnika označava da im se sviđa
    Poništi
  17. 12. stu 2018.

    2012: Using GPUs for Deep Learning 2018: Using Deep Learning for GPUs "Differentiable Monte Carlo Ray Tracing through Edge Sampling"

    5 proslijeđenih tweetova 53 korisnika označavaju da im se sviđa
    Poništi
  18. 2. kol 2018.

    OpenAI Five till Sunday:

    1 reply 5 proslijeđenih tweetova 44 korisnika označavaju da im se sviđa
    Poništi
  19. 3. srp 2018.

    Halfway through 2018, still my favorite lecture. One takeaway: Fish matrix: the matrix...but for fish.

    5 proslijeđenih tweetova 13 korisnika označava da im se sviđa
    Poništi
  20. 14. lip 2018.

    What I imagine most of my parameters are doing when training a (generative) adversarial network:

    9 proslijeđenih tweetova 76 korisnika označava da im se sviđa
    Poništi

@thisismyhat još nije objavio nijedan Tweet.

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:

    ·