Sergey Levine

@svlevine

Assistant Professor at UC Berkeley

Berkeley, CA
যোগদান করেছেন এপ্রিল ২০১৮

টুইট

আপনি @svlevine-কে ব্লক করেছেন

আপনি কি এই টুইটগুলি দেখতে চাওয়ার বিষয়ে নিশ্চিত? টুইটগুলি দেখা হলে @svlevine অবরোধ মুক্ত হবে না।

  1. পুনঃ টুইট করেছেন
    ২১ ঘন্টা আগে

    ICLR2019 will be in New Orleans💃🏾We just announced the call for papers with the submission deadline on 27 September. Happy to be one of this year's programme chairs with Alex Rush, & Karen Livescu. Don't forget to read the length instructions.

    পূর্বাবস্থায়
  2. ১৪ জুলাই

    Some more talks from our group on Sun at : ERL T1 11:30 Abhishek: meta-learning exploration PAL-18 C7 14:00 me: Five myths about deep RL - come find out what they are! 14:45 : model-based deep RL TADGM A5 9:50 : probabilistic MAML

    পূর্বাবস্থায়
  3. ১৪ জুলাই

    I'll also be speaking at nonconvex optimization (A6) on meta-learning, including universality, probabilistic meta-learning, and a discussion of our latest paper on probabilistic MAML (PLATIPUS):

    পূর্বাবস্থায়
  4. ১৪ জুলাই

    : Workshops are always my favorite part. Some talks from our group on Sat: GoalsRL A7 10:30 meta-IRL 10:40 visual RL LLARLA A4 9:20 14:55 Abhishek: unsupervised meta-learning AutoML K1 16:00 : meta-learning

    পূর্বাবস্থায়
  5. ২৯ জুন
    পূর্বাবস্থায়
  6. ২৯ জুন

    Deep RL from images that generalizes to new objects for robotic grasping and automatically learns a range of pregrasp manipulation strategies: See also

    পূর্বাবস্থায়
  7. পুনঃ টুইট করেছেন
    ২৮ জুন

    Meta-learning can enable a robot to imitate a vision-based skill from a single human demonstration! New BAIR blog post from Tianhe Yu and on "One-Shot Imitation from Watching Videos"

    পূর্বাবস্থায়
  8. ২৭ জুন

    Way to go Chelsea!! Some very well deserved recognition for "making robots act like toddlers" :)

    পূর্বাবস্থায়
  9. ২৬ জুন

    In 2017, there was a contest called "Learning to Run" at NIPS that produced gems like this: In 2018, it returns, but now with a prosthetic foot: objectives include "Deep Reinforcement Learning to solve problems in medicine"

    পূর্বাবস্থায়
  10. ২৬ জুন

    Come see our talks at : Wed 9:30: Dexterous Manipulation, Kumar*, Rajeswaran* et al. Thu 13:30: Shared Autonomy, Reddy et al. & One-Shot Imitation, Yu*, Finn* et al. Thu 14:30: my early career spotlight (a bit nervous for that one...)

    পূর্বাবস্থায়
  11. ২৬ জুন

    An interesting article comparing concerns about malicious AI with the behavior of current multinational corporations. Somehow reminds me of “Accelerando” by Stross...

    পূর্বাবস্থায়
  12. ১৩ জুন

    Meta-learning enables fast learning, but needs hand-engineered meta-training tasks. Can we get the tasks themselves automatically? Our first attempt at this for RL: unsupervised meta-reinforcement learning: w/ Abhishek Gupta, Ben Eysebach,

    পূর্বাবস্থায়
  13. পুনঃ টুইট করেছেন
    ৯ জুন

    Deep learning could improve their robot hugs by 23.7% or more! (Actually, we once used a machine-learning approach to detect whether a person was interested in a high five, fist bump, or hug from the robot and had it respond appropriately.)

    পূর্বাবস্থায়
  14. ৮ জুন

    Few-shot learning problems can be ambiguous. Now MAML can handle the ambiguity by sampling multiple classifiers via a Bayesian formulation of meta-learning. We call it PLATIPUS: w/ and some ambiguous celeba task:

    পূর্বাবস্থায়
  15. ১ জুন

    Reward design in RL is a bit of a mystical art. We can start to demystify it using control as inference, and also learn rewards from data using variational inverse control with events (VICE): w/ Justin Fu, Dibya Ghosh, Larry Yang

    পূর্বাবস্থায়
  16. ১ জুন

    Inverse RL + meta-learning: if you've inferred goals from observing other agents' actions in the past, you should be better at inferring new goals more quickly. Meta-learning seems like a natural fit for IRL: (w/ Ratner, Dragan, )

    পূর্বাবস্থায়
  17. পুনঃ টুইট করেছেন
    ৩১ মে

    Excited to announce our first class of AI Fellows, seven machine learning students to whom we’re collectively recommending $1.1 million in PhD fellowship support over the next five years:

    পূর্বাবস্থায়
  18. ৩১ মে

    Really nice article by Aaron Hertzmann about art and AI. Also a good read for anyone who wants a history lesson on art and technology:

    পূর্বাবস্থায়
  19. ৩১ মে

    Seems like really simple model-based RL methods can do just as well at convergence as model-free RL, if uncertainty is handled properly, with orders of magnitude fewer samples: with Kurtland Chua, , Rowan McAllister

    পূর্বাবস্থায়
  20. ২৩ মে

    Data-Efficient Hierarchical RL: Ant can now push blocks and “build” bridges by setting its own subgoals. With Ofir Nachum, S. Gu, and Honglak Lee

    পূর্বাবস্থায়

লোড হতে বেশ কিছুক্ষণ সময় নিচ্ছে।

টুইটার তার ক্ষমতার বাইরে চলে গেছে বা কোনো সাময়িক সমস্যার সম্মুখীন হয়েছে আবার চেষ্টা করুন বা আরও তথ্যের জন্য টুইটারের স্থিতি দেখুন।

    আপনিও পছন্দ করতে পারেন

    ·