Jonas Kubilius

@qbilius

What I cannot create, I do not understand. – R. Feynman

Vrijeme pridruživanja: travanj 2011.

Tweetovi

Blokirali ste korisnika/cu @qbilius

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

  1. Prikvačeni tweet
    5. ruj 2018.

    Super deep, feedforward DNNs might not be so lucrative to neuroscientists, so we built CORnet-S, a simple recurrent model mapped onto brain areas that is on par with the state-of-the-art on Brain-Score. Paper: , code+weights:

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

    Great talk today from and from lab . Recurrent connections important to explain visual processing in cortex.

    Poništi
  3. 12. pro 2019.

    Announcing Brain-Score, an integrative platform for comparing models to primate visual system, and CORnet-S, the current state-of-the-art model on Brain-Score. Learn more at our poster 190 at on Thursday morning. With

    Poništi
  4. proslijedio/la je Tweet
    12. pro 2019.

    Let's build a model of the brain! How? Come to our oral presentation at to find out -- we are presenting Brain-Score (quantifying match-to-brain) and CORnet-S (a shallow recurrent model that is more brain-like). Also streamed at . w/

    Prikaži ovu nit
    Poništi
  5. proslijedio/la je Tweet
    7. sij 2019.

    Great paper, but makes me really curious about how that Nature Machine Intelligence boycott is working out

    Prikaži ovu nit
    Poništi
  6. proslijedio/la je Tweet
    15. pro 2018.

    Excited to organize a workshop on future challenges for brain-like deep-nets with on March 04, 2019 in Cascais, Portugal. Our discussion panel could benefit from your discussions/feedback/questions. Check out more here:

    Prikaži ovu nit
    Poništi
  7. proslijedio/la je Tweet
    5. stu 2018.

    Deep artificial neural networks can predict the neural responses along the ventral visual pathway fairly well. They are imperfect yet useful. Find out how we used these networks to control the activity of single neurons in high level visual sensory cortex.

    Poništi
  8. proslijedio/la je Tweet
    4. stu 2018.

    is so full of random experiments and results — I don’t know who is going to put all of this together into a single model of anything..With every poster I am more and more convinced that () is the way to go ...

    Poništi
  9. proslijedio/la je Tweet
    21. ruj 2018.

    "Don't be scared of deep nets!" says scientists are using to understand the brain, covered by .

    Poništi
  10. proslijedio/la je Tweet
    9. ruj 2018.

    Here's our effort to make success in neuroscience quantifiable. Let's move beyond our favorite intuitions as "models", and build models that actually might resemble the brain and can be falsified!

    Poništi
  11. proslijedio/la je Tweet
    7. ruj 2018.

    Interested in brain-like neural networks at ? Stop by the Brain-Score and CORnet poster starting at 7:30 pm! w/

    Poništi
  12. proslijedio/la je Tweet
    7. ruj 2018.

    Five years ago, deep convolutional NNs inspired by brain science advanced to become the best models of the brain's visual processing. Since then, CV/ML has created more complex models, without regard to the brain. Here we aim to follow the brain science:

    Poništi
  13. proslijedio/la je Tweet
    5. ruj 2018.

    One of my dreams is to help accelerate progress in the brain sciences and AI by defining shared goals and facilitating ways to work toward them as a larger community of data collectors and model builders. We share a concrete step here called Brain-Score:

    Poništi
  14. proslijedio/la je Tweet
    5. ruj 2018.

    Very deep NNs are not a good match for ventral visual pathway activity. & CO build new ANNs with recurrent and skip connections + monitor both performance and the match between each new model and primate brain and behavioral data.

    Poništi
  15. proslijedio/la je Tweet
    5. ruj 2018.

    Two new papers on computational models of the brain: () is a composite benchmark quantifying brain-likeness of neural networks; CORnet () is a top model with a brain-like architecture w/

    Poništi
  16. 5. ruj 2018.
    Prikaži ovu nit
    Poništi
  17. 5. ruj 2018.

    Some co-author handles went missing:

    Prikaži ovu nit
    Poništi
  18. 5. ruj 2018.

    Ever wondered which deep nets were most brain-like? (spoiler: not the best ImageNet models) Check out and our latest paper ! Winners: DenseNet-169, ResNet-101 and... CORnet-S - read on...

    Prikaži ovu nit
    Poništi
  19. 11. kol 2018.

    Just heard of @NipsConference new name proposals: a NALS conference -- you thought NIPS sounded bad?; SNIPS -- appears to emphasize speedy snippets of short-term results; and ICLIPS -- the one directly leading to maximizing papers clips (). So maybe NILS?

    Poništi
  20. 26. srp 2018.

    Spent the whole day trying to figure out how to get the pid of a process executed remotely in the background. Why a process can't just return its pid is beyond me (), but the only way it works is ssh with Fabric, then run `nohup command &>/dev/null & echo $!`.Obviously!

    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:

    ·