Thomas Lahore

@evolvingstuff

Machine learning and occasionally some other stuff

Seattle, WA
Vrijeme pridruživanja: prosinac 2009.
202 fotografije ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @evolvingstuff

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

  1. 4. velj

    Radioactive data: tracing through training "makes imperceptible changes to this dataset such that any model trained on it will bear an identifiable mark. The mark is robust to strong variations such as different architectures or optimization methods"

    10 proslijeđenih tweetova 24 korisnika označavaju da im se sviđa
    Poništi
  2. proslijedio/la je Tweet
    30. sij

    Welcome OpenAI to the PyTorch community!

    We're standardizing OpenAI's deep learning framework on PyTorch to increase our research productivity at scale on GPUs (and have just released a PyTorch version of Spinning Up in Deep RL):
    164 proslijeđena tweeta 850 korisnika označava da im se sviđa
    Poništi
  3. proslijedio/la je Tweet
    28. sij

    Over a million particles running real-time on the gpu. They’re attracted to each other while having a weaker desire to reassemble the image.

    851 proslijeđeni tweet 4.587 korisnika označava da im se sviđa
    Poništi
  4. proslijedio/la je Tweet
    28. sij

    Finally, Differentiable Physics is Here! ▶️Full video (ours): 📜Source paper:

    48 proslijeđenih tweetova 127 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    26. sij

    Quaternions and Euler angles are discontinuous and difficult for neural networks to learn. They show 3D rotations have continuous representations in 5D and 6D, which are more suitable for learning. i.e. regress two vectors and apply Graham-Schmidt (GS).

    11 replies 147 proslijeđenih tweetova 663 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    23. sij

    20 million connections have been mapped between 25,000 neurons in the fruit fly brain by researchers at . “It’s a landmark,” says our Clay Reid who is working on a similar effort in the mouse brain.

    37 proslijeđenih tweetova 88 korisnika označava da im se sviđa
    Poništi
  7. proslijedio/la je Tweet
    22. sij

    The quiet semisupervised revolution continues

    FixMatch: focusing on simplicity for semi-supervised learning and improving state of the art (CIFAR 94.9% with 250 labels, 88.6% with 40). Collaboration with Kihyuk Sohn, Nicholas Carlini
    Prikaži ovu nit
    7 replies 221 proslijeđeni tweet 1.100 korisnika označava da im se sviđa
    Poništi
  8. proslijedio/la je Tweet
    20. sij

    Flax: A neural network library for JAX designed for flexibility (pre-release)

    11 replies 110 proslijeđenih tweetova 441 korisnik označava da mu se sviđa
    Poništi
  9. 20. sij

    Reformer: The Efficient Transformer "we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O(L^2) to O(L log L), where L is the length of the sequence" paper: code:

    1 reply 39 proslijeđenih tweetova 177 korisnika označava da im se sviđa
    Poništi
  10. proslijedio/la je Tweet
    17. sij

    Cool theory paper presenting a problem that: - can be efficiently learned by SGD with a DenseNet with x^2 nonlin, - cannot be efficiently learned by any kernel method, including NTK.

    1 reply 75 proslijeđenih tweetova 363 korisnika označavaju da im se sviđa
    Poništi
  11. proslijedio/la je Tweet
    16. sij

    ....The reason for this is also why it's more efficient for human engineers to build AI systems through machine learning than through direct programming. The price is training data.

    14 replies 24 proslijeđena tweeta 101 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  12. proslijedio/la je Tweet
    16. sij

    It is more efficient for evolution to specify the behavior of an intelligent organism by encoding an objective to be optimized by learning than by directly encoding a behavior. The price is learning time. The reason...

    30 replies 117 proslijeđenih tweetova 519 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  13. proslijedio/la je Tweet
    15. sij

    A fascinating new Nature paper from hypothesizes (and shows supporting data!) about how state of the art reinforcement-learning algorithms may explain how dopamine works in our brains.

    1 reply 52 proslijeđena tweeta 167 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  14. proslijedio/la je Tweet
    14. sij

    By restructuring math expressions as a language, Facebook AI has developed the first neural network that uses symbolic reasoning to solve advanced mathematics problems.

    849 proslijeđenih tweetova 2.581 korisnik označava da mu se sviđa
    Poništi
  15. proslijedio/la je Tweet
    13. sij

    🔥 Introducing Tokenizers: ultra-fast, extensible tokenization for state-of-the-art NLP 🔥 ➡️

    , , i još njih 3
    11 replies 234 proslijeđena tweeta 980 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  16. 13. sij

    What an elegant idea: Choosing the Sample with Lowest Loss makes SGD Robust "in each step, first choose a set of k samples, then from these choose the one with the smallest current loss, and do an SGD-like update with this chosen sample"

    5 replies 35 proslijeđenih tweetova 199 korisnika označava da im se sviđa
    Poništi
  17. proslijedio/la je Tweet
    11. sij

    With recent work that enables transformer to process very long training sequences, we could be only scratching the surface of the full capabilities of self-attention networks. They may have strong inductive bias to model things like hi-res video sequences.

    Reformer: The Efficient Transformer They present techniques to reduce the time and memory complexity of Transformer, allowing batches of very long sequences (64K) to fit on one GPU. Should pave way for Transformer to be really impactful beyond NLP domain
    Prikaži ovu nit
    9 proslijeđenih tweetova 68 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    11. sij

    On the Relationship between Self-Attention and Convolutional Layers This work shows that attention layers can perform convolution and that they often learn to do so in practice. They also prove that a self-attention layer is as expressive as a conv layer.

    i
    7 replies 278 proslijeđenih tweetova 1.132 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  19. 8. sij

    DiffTaichi: Differentiable Programming for Physical Simulation "a new differentiable programming language tailored for building high-performance differentiable physical simulators"

    8 proslijeđenih tweetova 15 korisnika označava da im se sviđa
    Poništi
  20. proslijedio/la je Tweet
    6. sij

    Here is how AI ate the keyboard

    175 replies 5.023 proslijeđena tweeta 9.910 korisnika označava da im se sviđa
    Poništi

@evolvingstuff 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:

    ·