MegaExponential TF

@furlanel

Employers and friends have no responsibilities for my thoughts.

Vrijeme pridruživanja: lipanj 2016.

Tweetovi

Blokirali ste korisnika/cu @furlanel

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

  1. Prikvačeni tweet
    26. lip 2019.

    New pre-print on learning task agnostic representations for partially observable environments in reinforcement learning. With Amy Zhang Luis Pineda Laurent Itti and Joelle Pineau

    Poništi
  2. proslijedio/la je Tweet
    17. srp 2019.

    Not demonstrated is stimulation, "writing" activity. Stim causes electrical surge that will briefly "blind" recording. How long does the artifact last? How many neighboring threads will be blinded? Important Qs for bi-directional BCI.

    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet

    So, it looks like have quickly caught up with the incumbents (e.g. blackrock). There is nothing revolutionary but a range of really creative ideas (love the guide tube like needle idea). I need to see more data but they seem to be on a great track.

    Prikaži ovu nit
    Poništi
  4. proslijedio/la je Tweet
    16. srp 2019.

    Just returned from two days in SF reviewing grants for the NIH Small Business: Neuroscience Assay, Diagnostics and Animal Model Development panel. I'm warmed up for 's event tonight. At 8pm I'll live-tweet the event with my first-take "review" of the tech.

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    30. lip 2019.

    Google and the University of Chicago were recently sued over sharing patient data without correct de-identification and consent FYI exact times should be removed from data before any sharing, but that's not even all the problem 1/3

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    29. lip 2019.

    La forza di chi urla contro è la forza della disperazione di chi in fondo sa di stare dalla parte fallita della storia ... sono quelli che 50 anni fa applaudivano la polizia che irrompeva a Stonewall Inn ...

    Poništi
  7. proslijedio/la je Tweet
    27. lip 2019.

    Learning Causal State Representations of Partially Observable Environments w/ Self-supervised method for learning a minimal sufficient statistic for next step prediction, show systematic improvement of DQN & tabular model

    Poništi
  8. proslijedio/la je Tweet
    26. lip 2019.
    Poništi
  9. proslijedio/la je Tweet

    Astrocytes Contribute to Remote Memory Formation by Modulating Hippocampal-Cortical Communication During Learning

    Poništi
  10. proslijedio/la je Tweet

    Real world scenarios for reinforcement are partially observable, unlike atari games. Causal states optimally compress the joint history of actions and observations in a task agnostic manner. Amy Zhang Joelle Pineau

    Poništi
  11. proslijedio/la je Tweet
    24. svi 2019.

    Send me an essay describing where you want to go this summer and why. Open to undergraduate women from anywhere in the world. Please keep your essay < 1000 words, email to dortsao@caltech.edu by June 10, 2019 Please retweet

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet

    Come to our poster on learning minimal sufficient representations in partially observable decision process at at 5pm at partial observability workshop

    Poništi
  13. proslijedio/la je Tweet

    To help reviewing neophytes struggling to cope w the deluge of ML papers, a very special post: ***Is This a Paper Review?***

    Poništi
  14. proslijedio/la je Tweet
    28. lis 2018.
    Poništi
  15. 28. lis 2018.

    Maintaining my constant support (in real life first) for any minority and person that is treated with any lack of kindness. this is just mean. The discussion of policies' effect and consequent opportunity costs is fundamental. Attention is limited.

    Poništi
  16. proslijedio/la je Tweet
    27. lis 2018.

    if all those "small data" in medical imaging from all those talks, which end unanimously with the problem of small data in medical imaging, were shared and merged, I don't think there would be a problem with small data in medical imaging.

    Poništi
  17. proslijedio/la je Tweet
    25. lis 2018.
    Poništi
  18. proslijedio/la je Tweet

    Well that's funny. We could make a model that mapped from latent vectors R^n -> R^n with no compression and full rank operators. Would we then own all of R^n?! Let's do it! I hereby declare myself Regent of the Reals!!

    Poništi
  19. proslijedio/la je Tweet
    Odgovor korisniku/ci

    "Use your whole keyboard to control a spaceship with modules for speed/attack/defense that you can assign for every key input When you get dammage, you lose key inputs" project died but I always say It'll be alive again someday

    Poništi
  20. proslijedio/la je Tweet
    14. lis 2018.
    Odgovor korisniku/ci

    ML results are definitely not hardware agnostic. You should expect to need to retune hyperparameters at least a little bit whenever you change any part of the hardware / software infrastructure that changes the patterns of rounding error.

    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:

    ·