Bored Yann LeCun

@boredyannlecun

Musing on the rise of deep learning in Yann's downtime. (parody) email: boredlecun@gmail.com

Vrijeme pridruživanja: ožujak 2016.

Tweetovi

Blokirali ste korisnika/cu @boredyannlecun

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

  1. Prikvačeni tweet
    19. lis 2018.

    Many people are not aware that most elementary operations on all numbers (multiplication & addition) are simply special cases of (1x1) convolutions. Let's finally start to teach convolution in its full glory to all children from primary school!

    Poništi
  2. 27. sij

    For all their marvels, no ConvNet could survive training 1% as hard as Kobe Bryant. RIP

    Poništi
  3. proslijedio/la je Tweet
    Poništi
  4. 4. sij

    In 2020, I predict we sill see the first ***ConvPuter***—a fully-functioning computer system where every function in every application, from the OS kernel to user-facing apps, is learned w. deep convolutional neural networks.

    Poništi
  5. proslijedio/la je Tweet
    2. sij
    Odgovor korisniku/ci

    Or "The 10's", has a nice ring to it

    Poništi
  6. 2. sij

    Still no consensus on what to call the 2010s. I'm personally partial to "CONVNETS FOR THE WIN", but open to other, similarly-thoughtful suggestions mentioning ConvNets in caps.

    Poništi
  7. 30. pro 2019.

    Used to drop the loss But now I'm the boss So drink champagne & listen to Don Cherry 🎺🎧 2010s were the cake, 2020s the cherry May your new year be merry... ML Twitter Try not to let make it too bitter.

    Poništi
  8. 30. pro 2019.

    You want to fix our broken society? How about swapping the House of Representatives for a ***House of Representations***?. You w. me ? ?

    Poništi
  9. proslijedio/la je Tweet
    20. pro 2019.

    You can tune a hyperparameter, but you can't tuna fish.

    Poništi
  10. 17. pro 2019.
    Poništi
  11. proslijedio/la je Tweet
    15. pro 2019.

    “For the 10yrs CVPR wouldn’t publish neural net papers, I made collages out of rejected drafts” — (?) Enjoy this thread on the flight back home from 🤣🤖👇🏻

    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    15. pro 2019.

    This is the price I pay for cancelling an invited talk at a workshop on Machine Learning for Creativity and Design, when one of the organizers is an improv performer and...

    Poništi
  13. proslijedio/la je Tweet
    15. pro 2019.

    We did a delightful exercise in collaborative comedy writing and improv performance b/w + and the mysterious and hilarious ...

    Poništi
  14. proslijedio/la je Tweet
    15. pro 2019.
    Poništi
  15. proslijedio/la je Tweet
    15. pro 2019.
    Odgovor korisnicima

    Thanks for filling for me guys.

    Poništi
  16. 14. pro 2019.

    ***Towards Sustainable Architecture: 3D Convolutional Neural Networks for Computational Fluid Dynamics Simulation and Reverse Design Workflow*** The length of this title may seem daunting, but fortunately, you can handle it using ConvNets with mean pooling.

    Prikaži ovu nit
    Poništi
  17. 14. pro 2019.

    ***First Steps Towards Collaborative Poetry Generation*** This collaboration is a nice idea, but note that the finest poetry ever flows from that storied collaboration between the 2D convolution and the max-pooling operation.

    Prikaži ovu nit
    Poništi
  18. 14. pro 2019.

    ***Artist in the Cloud: Towards an Autonomous Artist*** At first, I thought “artist in the cloud” was a paper about Twitter parody accounts. Despite my disappointment, I will try to attend this poster.

    Prikaži ovu nit
    Poništi
  19. 14. pro 2019.

    ***Towards a Principled Evaluation of Likeability for Machine-Generated Art*** On the topic of popularity contests, the convolutional neural network was recently named by Lyon Magazine to be the #1 most popular invention in France of the last 50 yrs.

    Prikaži ovu nit
    Poništi
  20. 14. pro 2019.

    ***Dialog on a Canvas with a Machine*** This work helps humans to expand their creative capacities. I will not comment on how I expanded my creative capacities in the 1970s, but let’s just call it “convolution”.

    Prikaži ovu nit
    Poništi
  21. 14. pro 2019.

    ***Neural Loops*** This work reminds us that the passage of time itself is 1D convolution.

    Prikaži ovu nit
    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:

    ·