Xander Steenbrugge

@xsteenbrugge

Machine Learning researcher & online educator. Passionate about AI, Virtual Reality, renewable energy, space exploration, digital art & philosophy.

België
youtube.com/arxivinsights
Vrijeme pridruživanja: siječanj 2015.
134 fotografije ili videozapisa Fotografije i videozapisi

Tweetovi

Blokirali ste korisnika/cu @xsteenbrugge

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

  1. Prikvačeni tweet
    28. lis 2019.

    Emtremely excited to finally share a side-project I've been working on for the past few months: Neural Synesthesia, visualizing music with ! So many ideas left to explore, this journey is just getting started.. More HD renders on my new YT channel:

    71 reply 1.126 proslijeđenih tweetova 3.532 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    4. velj

    New Paper: Adversarial Generation of Continuous Implicit Shape Representations We propose two GANs with a DeepSDF network as the generator and either a 3D CNN or a Pointnet as the discriminator. Written by and me! Paper:

    6 replies 34 proslijeđena tweeta 179 korisnika označava da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  3. 2. velj

    GAN-idea: 1. Train a VAE with strong Gaussian prior (Decoder = StyleGAN's G architeture, Encoder = D, but final layer maps to 512 outs) 2. Init StyleGAN with these nets and fine-tune under normal GAN framework. - Would this avoid mode collapse? - Any prior work like this?

    10 replies 20 proslijeđenih tweetova 159 korisnika označava da im se sviđa
    Poništi
  4. proslijedio/la je Tweet
    31. sij

    A full build of Autopilot neural networks involves 48 networks that take 70,000 GPU hours to train. Together, they output 1,000 distinct tensors (predictions) at each timestep. This is what a Tesla autopilot sees [source: ]

    45 replies 1.082 proslijeđena tweeta 3.380 korisnika označava da im se sviđa
    Poništi
  5. 1. velj

    Neural nets are universal function approximators. While powerful, this also means NNs will fit any training corpus without generalization guarantees. Engineering the agent's perception of the surrounding world could be a strong prior to encourage learning generalizable skills!

    To the left, you see a trained agent playing a level of a game. To the right, you see the same playthrough from an agent-centric perspective: cropped, translated, and rotated with the agent in the center. Which perspective is the best input for the agent?
    Prikaži ovu nit
    1 reply 8 proslijeđenih tweetova 32 korisnika označavaju da im se sviđa
    Poništi
  6. proslijedio/la je Tweet
    30. sij

    Full body characters for Artbreeder are finally live!! Wild, weird and very diverse - you can find your new character ideas here. So much has already been created in the first few hours.

    11 replies 42 proslijeđena tweeta 173 korisnika označavaju da im se sviđa
    Poništi
  7. proslijedio/la je Tweet
    25. sij

    +1! IMHO the most successful machine learning system of all, internet search, is a great example of this. search is just a super powered externalised memory. the next-big-thing (TM) will be super powered externalised cognition. we want to augment ppl rather than automating them.

    Honestly, the question is not, and has never been, "can machine learning replace radiologists/etc" (which won't happen in the foreseeable future). The question is, how can radiology/etc utilize ML to improve outcomes, decrease the cost of care, and broaden accessibility.
    Prikaži ovu nit
    14 proslijeđenih tweetova 88 korisnika označava da im se sviđa
    Poništi
  8. proslijedio/la je Tweet
    22. sij

    Cursed StyleGAN2 people after a deep style transfer, makes some interesting things

    29 proslijeđenih tweetova 132 korisnika označavaju da im se sviđa
    Poništi
  9. 24. sij

    I rather enjoy the geometrical patterns of this latest checkpoint! It's trying to fit the regularities & symmetries of the training imgs, getting close and failing 'organically'.. Like an uncanney valley for geometrical shape 👌😋

    16 proslijeđenih tweetova 120 korisnika označava da im se sviđa
    Poništi
  10. proslijedio/la je Tweet
    18. sij

    Amazing frame interpolation by DAIN: Depth Aware Video Frame Interpolation.

    39 replies 1.147 proslijeđenih tweetova 4.371 korisnik označava da mu se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  11. 18. sij

    This is incredible work! After all of the compute-heavy autoregressive audio models, this feels like it could finally make AI-based electronic music composition a reality. How shall we call the new, emerging genre.. mAIsic? 🤔😅 Going to play with this big-time 👏🎶

    Differentiable Digital Signal Processing (DDSP)! Fusing classic interpretable DSP with neural networks. ⌨️ Blog: 🎵 Examples: ⏯ Colab: 💻 Code: 📝 Paper: 1/
    Prikaži ovu nit
    3 proslijeđena tweeta 41 korisnik označava da mu se sviđa
    Poništi
  12. 18. sij

    Just stumbled upon this wonderful footage of a discussion at Yoshua Begio's lab about generalizing the concept of backpropagation. This kind of intelligent, constructive brainstorming really demonstrates the power of AI research labs! Amazing 👏👨‍🎓

    1 reply 13 proslijeđenih tweetova 68 korisnika označava da im se sviđa
    Poništi
  13. 17. sij

    Just got Live, GAN-based visuals up and running in ! Anybody have a gig coming up and need some kick-ass visuals? 👌😎 More full renders on or on Instagram: @neurasynth

    8 replies 27 proslijeđenih tweetova 130 korisnika označava da im se sviđa
    Poništi
  14. proslijedio/la je Tweet
    14. sij

    BigGAN samples are famously photo-realistic but limited in diversity for some classes. Slightly modifying only the class embeddings (network unchanged) can reduce the diversity gap by ~50%! Work with Long Mai and led by fantastic !! Paper & video:

    67 proslijeđenih tweetova 243 korisnika označavaju da im se sviđa
    Prikaži ovu nit
    Prikaži ovu nit
    Poništi
  15. proslijedio/la je Tweet
    14. sij

    Can Giraffes Become Birds? An Evaluation of Image-to-image Translation for Data Generation pdf: abs:

    28 proslijeđenih tweetova 92 korisnika označavaju da im se sviđa
    Poništi
  16. 11. sij

    First LSTM's and now even CNNs, it does seem like attention really is all we need.. Applying these attention mechanisms to vision models could open up a totally new area of "explainable AI" models 👌

    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.
    Prikaži ovu nit
    1 reply 74 proslijeđena tweeta 332 korisnika označavaju da im se sviđa
    Poništi
  17. proslijedio/la je Tweet
    4. sij

    Metabolic Engineering and Synthetic Biology of Yeast - Jens Nielsen🤯 (the whole channel is quite good). Bio will grow into a major tech stack. We're writing assembly today, but when the AWS is up things will get interesting.

    12 replies 96 proslijeđenih tweetova 564 korisnika označavaju da im se sviđa
    Poništi
  18. proslijedio/la je Tweet
    26. pro 2019.

    I am so excited for when drops Vid2Game code. Imagine all the amazing things you could do with this. Maybe the models are even playable in a Colab? FMV games: The Next Generation? abs: vid:

    7 replies 106 proslijeđenih tweetova 414 korisnika označava da im se sviđa
    Poništi
  19. proslijedio/la je Tweet
    25. pro 2019.

    If you want to be rewarded, you have to be irreplaceable. If you want to be irreplaceable, you have to be unique. If you want to be unique, you have to be authentic. If you want to be authentic, stop listening to everyone and everything else. It’s drowning “you” out.

    10.305 proslijeđenih tweetova 40.375 korisnika označava da im se sviđa
    Poništi
  20. 24. pro 2019.

    Now here's another StyleGAN use I hadn't thought of, brilliant! Perhaps can use this to design the next sci-fi model 👌😅

    1:00
    wip LSUN Car to concepts
    Prikaži ovu nit
    6 proslijeđenih tweetova 65 korisnika označava da im se sviđa
    Poništi
  21. proslijedio/la je Tweet
    20. pro 2019.

    How dimensional ultrasound scans and computer graphics get these impressive images of baby animals in the womb

    8 replies 68 proslijeđenih tweetova 300 korisnika označava da im se sviđa
    Poništi

@xsteenbrugge 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:

    ·