Durk Kingma

@dpkingma

Research scientist in the field of machine learning, at Google Research / Brain Team. Ph.D. from Univ. of Amsterdam.

Vrijeme pridruživanja: ožujak 2009.

Tweetovi

Blokirali ste korisnika/cu @dpkingma

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

  1. Prikvačeni tweet
    30. ruj 2017.

    "Variational Inference and Deep Learning: A New Synthesis", written by yours truly, is now available for D/L here: .

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    6. velj

    Organizers of data science and machine learning conferences (NeurIPS, ICML, AISTATS, ICLR, UAI, ...): Allow remote paper & poster presentations at conferences - Sign the Petition! via

    Prikaži ovu nit
    Poništi
  3. 29. sij

    The next decades are going to be very interesting as machine learning for robotics *really* starts to positively impact our economy. Congrats to for this launch, and robotics friends , , for pushing the envelope.

    Poništi
  4. 27. sij

    Another great result demonstrating that VAEs (deep learning + amortized variational inference) make a lot of sense for data compression. Its loss function directly maximizes compressibility, and the resulting codec is fully parallelizable.

    Poništi
  5. 8. sij

    Our paper "Variational Autoencoders and Nonlinear ICA: A Unifying Framework" has been accepted to AISTATS'20. With , Ricardo Pio Monti and Aapo Hyvarinen (UCL). Surprisingly strong and general identifiability results, with rigorous proofs!

    Poništi
  6. 26. pro 2019.

    Clearly, symbolically defining the term "deep learning" is futile, and we can only accurately represent the concept with a deep neural net.

    Poništi
  7. 13. pro 2019.

    Interested in scalable algorithms for learning energy-based models? Check out our preprint on Flow Contrastive Estimation (FCE): an elegant, adaptive extension of noise-contrastive estimation, with really promising results! Work led by brilliant .

    Poništi
  8. proslijedio/la je Tweet
    11. pro 2019.

    Our work "Flow Contrastive Estimation of Energy-Based Models" (, , , Z. Xu , YN Wu) at : We show (1) joint learning of EBM & Glow, (2) correction of over-dispersion, (3) competitive semi-super cls.

    Prikaži ovu nit
    Poništi
  9. proslijedio/la je Tweet
    4. pro 2019.
    Poništi
  10. proslijedio/la je Tweet

    I'd like to organize a "manifesto track" at AI conferences where speakers come up and present their controversial, slightly unhinged views on how to achieve AGI. Experimental evidence is optional. - Numenta HTM - LeCake - - Non-Axiomatic Reasoning System

    Prikaži ovu nit
    Poništi
  11. 19. ruj 2019.
    Prikaži ovu nit
    Poništi
  12. 19. ruj 2019.

    Ever wondered, like me, if you can can turn hierarchical VAEs into a fully parallelizable lossless compression scheme? Wonder no more: new blogpost by Friso on "Bit-Swap", his recursive algorithm for doing just that: Disclaimer: he's my sibling 🤓

    Prikaži ovu nit
    Poništi
  13. 13. ruj 2019.

    . Some pretty fresh tunes generated with the help of AI . I actually like these tunes a lot! Well crafted, and collaborators. More info:

    Poništi
  14. 6. ruj 2019.

    ... And thanks to for being such a dynamic and inspiring place for doing research! (Hint: apply for positions at, for example, and the lab in Amsterdam)

    Prikaži ovu nit
    Poništi
  15. 6. ruj 2019.

    It's been 2 years since I finished my PhD thesis, but it won the ELLIS PhD award! All thanks to for being possibly the world's best PhD advisor & thanks to great collaborators , , , Jimmy Ba, etc.! [1/2]

    Prikaži ovu nit
    Poništi
  16. 30. kol 2019.

    Someone is obviously really close to solving AGI:

    Poništi
  17. 20. srp 2019.

    Follow Apollo 11 in real time:

    Poništi
  18. proslijedio/la je Tweet
    12. srp 2019.

    Nonlinear ICA is known as unidentifiable with data only. They show that VAE conditioned with additional observation (e.g., time index, class label) can identify true components up to trivial transformation, showing principled disentanglement ability of VAE

    Poništi
  19. proslijedio/la je Tweet
    30. lip 2019.

    Dear friends, I’m extremely excited for our current research for “Latent Cinema” idea. This piece is narrated from ~300k collective photographic memories of Stockholm. Our GAN browser practically opens a camera into the space in the mind of a machine.

    Poništi
  20. proslijedio/la je Tweet
    21. lip 2019.

    Excited to share our latest work; it combines neural networks with stochastic processes by positing priors over the relational structure of the dataset, without relying on global latent variables!

    Poništi
  21. 18. lip 2019.

    "In addition [...] we will give $50 million in grants through to nonprofits focused on the issues of homelessness and displacement. This builds on the $18 million in grants we’ve given to help address homelessness over the last five years [...]." 👌

    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:

    ·