Giuseppe Cuccu

@giuse_tweets

Senior Machine Learning researcher . Addressing RL continuous control problems by improving neuroevolution.

Fribourg, Switzerland
exascale.info/giuse
Vrijeme pridruživanja: svibanj 2017.
12 fotografija ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @giuse_tweets

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

  1. Prikvačeni tweet
    30. lip 2019.

    New evolutionary algorithm almost done: here a four-neurons network plays bipedal walker in just a couple of hours of neuroevolution on an old workstation, no gpus :)

    11 replies 52 proslijeđena tweeta 328 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    6. sij

    I'm very happy to announce that I will receive the IEEE CIS Early Career Award 2020. The citation is "For Contributions to Computational Intelligence and Games".

    Professor honored with Early Career Award
    24 proslijeđena tweeta 450 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    17. lis 2019.

    Neural architecture evolution, a new method? Sounds pretty similar to neuroevolution to me.

    Neural architecture evolution, a new method for automatically finding optimal neural networks for video understanding, has yielded architectures that outperform existing hand-made models and show improvements to network runtime of 10-100x. Learn more ↓
    10 replies 24 proslijeđena tweeta 116 korisnika označava da im se sviđa
    Poništi
  4. proslijedio/la je Tweet
    12. lis 2019.

    Tokyo, minus people. Photos taken immediately post-typhoon.

    台風で人がいなくなった新宿。かなり異世界感あって撮影楽しかった。
    Prikaži ovu nit
    14 replies 630 proslijeđenih tweetova 1.896 korisnika označava da im se sviđa
    Poništi
  5. proslijedio/la je Tweet
    6. lis 2019.

    Why is machine learning rarely used for generating game levels? Because game levels are harder to generate than, say, faces. And because there aren't enough levels to train on. HOWEVER! In our new paper, we learn to generate pretty good levels after training on only 5 levels!

    7 replies 135 proslijeđenih tweetova 512 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    27. ruj 2019.

    I've been using this terminal app called thefuck for years and it still cracks me up because I'm 12. It corrects errors in previous console commands. So if you type the wrong thing, you just say 'fuck' and it gives you what you probably meant.

    366 replies 3.195 proslijeđenih tweetova 16.363 korisnika označavaju da im se sviđa
    Poništi
  7. proslijedio/la je Tweet
    25. ruj 2019.

    You can reject a paper but you can’t reject an idea 💡

    Odgovor korisniku/ci @hardmaru
    No opinion on favorite or not, but this paper , , & I submitted to NeurIPS'14 was rejected (~2K citations): Distilling the Knowledge in a Neural Network 2/3 said "1: This work is incremental and unlikely to have much impact"
    20 proslijeđenih tweetova 191 korisnik označava da mu se sviđa
    Poništi
  8. proslijedio/la je Tweet
    12. ruj 2019.

    We have a PhD scholarship in CS (), for someone interested in modelling/analysing incentives in Science and the production of knowledge. Part of an interdisciplinary project funded by . See details here:

    1 reply 18 proslijeđenih tweetova 16 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    11. ruj 2019.

    If funding is to play a role in hiring/tenure/promotion, it should be as a denominator. Divide your research accomplishments by the funding you got. You'd rather hire someone who can get amazing research done on a shoestring than someone who gets lots of money and squanders it.

    A number I'd find interesting: take your favorite research output metric, and normalize it by research group funding (salaries+grants+etc). Who's producing high levels of research-per-dollar? My guess is this looks totally different than the non-normalized research rankings.
    Prikaži ovu nit
    6 replies 15 proslijeđenih tweetova 57 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    27. kol 2019.

    Weight, weight: don't tell me.

    Can neural network architectures alone, without learning any weight parameters, encode solutions for a given task? We search for “weight agnostic neural network” architectures that can perform various tasks even when using random weight values. Learn more→
    7 replies 159 proslijeđenih tweetova 906 korisnika označava da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    22. kol 2019.
    Odgovor korisniku/ci

    there is actually a whole active area of research within the field of neurevolution on evolving the "genetic bottleneck," which is called in ML an *indirect encoding*. See for example HyperNEAT: ...(more)

    1 reply 8 proslijeđenih tweetova 43 korisnika označavaju da im se sviđa
    Poništi
  12. proslijedio/la je Tweet
    25. kol 2019.

    the hong kong police so far have: -tear gassed innocent people -dropped their guns -fired their guns in the air -arrested children -completely failed to change or control the situation in hong kong in any way

    And now a 12-year-old has been arrested.
    19 replies 96 proslijeđenih tweetova 165 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  13. 6. kol 2019.

    [ ]

    3 korisnika označavaju da im se sviđa
    Poništi
  14. proslijedio/la je Tweet
    30. srp 2019.

    There are 12 postdoc positions available in various fields (including explainable AI, AI in games, deep learning) as part of a collaboration between different Copenhagen universities. Please RT. Let's work together!

    76 proslijeđenih tweetova 69 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    16. srp 2019.

    Can your RL agent solve this level?

    0:59
    I hate this. I hate this like I haven’t hated anything for years. I hate this like I was born to hate this.
    Prikaži ovu nit
    21 reply 70 proslijeđenih tweetova 409 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  16. proslijedio/la je Tweet
    4. srp 2019.

    the struggle is real

    11 replies 413 proslijeđenih tweetova 1.479 korisnika označava da im se sviđa
    Poništi
  17. proslijedio/la je Tweet
    3. srp 2019.

    "python is the second best language for everything" -

    Python is much more than data analysis/modeling too. Useful for web dev & APIs, video games, random scripts. It's the second best language for everything. Also, comprehensions are the best.
    Prikaži ovu nit
    32 proslijeđena tweeta 126 korisnika označava da im se sviđa
    Poništi
  18. 3. srp 2019.

    Once again huge thanks to for their gym environment! Your contribution is invaluable, thanks guys! <3

    1 proslijeđeni tweet 2 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. 3. srp 2019.

    Actually since velocities are provided it's possible a FFNN would suffice, while BipedalWalkerHardcore seems to need an hidden layer. I am currently extending `tinynet` to both features.

    1 reply 1 proslijeđeni tweet 3 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  20. 3. srp 2019.

    Ok here are the pre-trained weights and the ~20 lines of code you need to make the video -- remember to `pip install` tinynet and gym, and install the `swig` package required by box2d games: ( may be interested?)

    1 proslijeđeni tweet 10 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  21. 2. srp 2019.

    The hardcore bipedal is more complex, I guess I will need to update my tiny neural network class to be multilayer :P => `pip install tinynet`

    1 proslijeđeni tweet 3 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi

@giuse_tweets 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:

    ·