Ian Osband

@IanOsband

Research scientist at DeepMind working on decision making under uncertainty. All tweets and views are mine alone.

Vrijeme pridruživanja: srpanj 2012.

Tweetovi

Blokirali ste korisnika/cu @IanOsband

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

  1. Prikvačeni tweet
    13. kol 2019.

    Really excited to release to the public! - Clear, scalable experiments that test core capabilities. - Works with OpenAI gym, Dopamine. - Detailed colab analysis - Automated LaTeX appendix Example report:

    Poništi
  2. proslijedio/la je Tweet

    The field is already self-correcting. Good departments/labs are clearing their eyes, caring less about paper count, seeing through the noise. Don't worry so much about the ICML deadline. Slow down, relax, try to do work you're proud of, submit when it's ready.

    Poništi
  3. 23. sij

    "One weird trick" for DQN in large (continuous) action spaces: - Initialize uniform action-sampling distribution. - Choose sampled action with highest Q. - Train sampling to produce "best action" + also some entropy. - ... Works surprisingly well! Great stuff , !

    Poništi
  4. 21. sij

    Ongoing investigation suggests this may have been originally destined for the Google kitchen "Labyrinth"... How I became custodian of this ham still remains a mystery.

    Prikaži ovu nit
    Poništi
  5. 20. sij

    A giant bag of ham has been delivered to my work addressed to me, and I don't know why... Where do I go from here?

    Prikaži ovu nit
    Poništi
  6. 19. sij

    Thank you to for organizing this truly inter-disciplinary conference: Curiosity, Creativity, and Complexity ... Looks like it's going to be very interesting!

    Poništi
  7. 13. sij

    Thought-provoking book, thanks : The Order of Time TL;DR: Time as we know it (fundamentally ordered from past to future) does not exist. Our perception of time is a side-effect of us residing in a low-entropy region of space + 2nd law.

    Poništi
  8. proslijedio/la je Tweet

    Run back Facebook did today. They bought a story in a publication, pretended it wasn’t sponsored, their COO posted it, then when they got caught, they denied knowing anything about it. And what was that ad/story about? Their commitment to stopping disinformation. Incredible.

    Prikaži ovu nit
    Poništi
  9. 7. sij
    Poništi
  10. proslijedio/la je Tweet
    Odgovor korisnicima i sljedećem broju korisnika:

    so you mean perhaps people should read the paper *first* before posting angry messages with incorrect information on twitter?

    Poništi
  11. proslijedio/la je Tweet
    28. pro 2019.

    It would help the discussion if everyone first 1. reads a causal inference book, eg , 2. watches a deep learning course emphasising modularity, compositionality and automatic differentiation, 3. implements the CI book examples in eg

    Poništi
  12. proslijedio/la je Tweet
    22. pro 2019.
    Poništi
  13. proslijedio/la je Tweet
    23. pro 2019.
    Odgovor korisnicima i sljedećem broju korisnika:

    Maxent policy gradient etc are very valuable RL algorithms, but many people seem confused by terms like posterior / optimal as used in RL as inference - they make sense in that context but some think it is truly Bayesian and therefore already handles exploration 'for free'.

    Poništi
  14. proslijedio/la je Tweet
    23. pro 2019.

    Medium also needs your clothes, your boots, and your motorcycle.

    Poništi
  15. 22. pro 2019.

    I actually don't think this is controversial... And I'm definitely "team Bayes" Yes, an independent Gaussian prior over NN weights is nonsense... We know the *interaction* is the most important part! But there's still huge potential for effective Bayesian deep learning!

    Poništi
  16. 21. pro 2019.

    Have you heard of "RL as Inference"? ... you might be surprised that this framing completely ignores the role of uncertainty! (confusing, since it talks a lot about "posteriors") Our spotlight tries to make sense of this:

    Poništi
  17. 20. pro 2019.

    Just waited 1hr for that estimated 20min delivery (underestimate by design) But the best part is no driver or food actually arrived and the app doesn't refund you without waiting for customer service... I think I'll be your newest customer!

    Poništi
  18. proslijedio/la je Tweet

    Zac Goldsmith, what a story. Son of humble billionaire, at 23 made editor of a magazine (owned by his uncle), gets Tory nom for leafy London seat at 32, loses the seat twice in 3 years, (also losing London mayoral race). Elevated to the House of Lords. Truly, the British dream.

    Poništi
  19. 15. pro 2019.

    This is why you need : It's under the hood but uses a "grammar of graphics" that copies 's from R... Almost like to Seriously only takes 1 day to get up to speed... You will not regret it.

    Poništi
  20. 14. pro 2019.

    Great talk from Ben Van Roy at the workshop on optimization for RL. Is it time for the field to move beyond "MDP"? Thinking about "agent state" might be a better perspective for learning in complex worlds... the real world "state" is just too complex!

    Poništi
  21. 14. pro 2019.

    Huge congratulations to and team... Definitely one of the biggest results in AI research this year!

    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:

    ·