Jean-Rémi King

@jrking0

Researcher. Human Intelligence, Neuroimaging and Machine Learning

Vrijeme pridruživanja: ožujak 2010.

Tweetovi

Blokirali ste korisnika/cu @jrking0

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @jrking0

  1. Prikvačeni tweet
    19. stu 2019.

    "The Human Brain encodes a Chronicle of Visual Events at each Instant of Time", by and I: the tl;dr thread:

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    8. sij

    Interactive visualization of brain activity with is getting an update thanks to and (to appear soon in the next MNE release and still WIP)

    Poništi
  3. proslijedio/la je Tweet
    5. pro 2019.

    If you are an academic and you travel to conferences etc please take a moment to fill this in - we are very interested in what climate-related policies would attract most support 🙏

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

    Pour tous ceux qui s'intéressent aux travaux de notre Conseil Scientifique, ce livre "La science au service de l'école" (coédition Canopé/Odile Jacob) passe en revue les actions du CSEN depuis près de deux ans: lecture, évaluations, métacognition, etc

    Poništi
  5. proslijedio/la je Tweet
    25. stu 2019.

    Applications for the Facebook AI Residency program are open. US (NYC, Seattle, Menlo Park): UK (London): Deadline: 2020-01-31

    Poništi
  6. proslijedio/la je Tweet
    27. stu 2019.

    Our work on 'Distributed coding of choice, action, and engagement across the mouse brain' is published today in Nature! With , , and . Here's a quick recap and what's new since bioRxiv.

    Prikaži ovu nit
    Poništi
  7. 27. stu 2019.

    (Put the cell in markdown to go back to default.)

    Prikaži ovu nit
    Poništi
  8. 27. stu 2019.

    Execute this in a code cell 2/2: %%html <style> -container {width: 100%; background-color: } .code_cell {flex-direction: row !important;} .code_cell .output_wrapper {width: 50%;background-color: } .code_cell .input {width: 50%;background-color: } </style>

    Prikaži ovu nit
    Poništi
  9. 27. stu 2019.

    Jupyter: improve your notebook workflow with a simple trick. I often get asked, so twitter might find it useful. 1/2

    Prikaži ovu nit
    Poništi
  10. proslijedio/la je Tweet
    26. stu 2019.

    New preprint👉 a) complementary role of MEG & fMRI for prediction and brain-behavior association b) MEG highlights source power in beta-alpha range c) multimodal prediction always pays off even with missing data

    , , i još njih 6
    Poništi
  11. 19. stu 2019.

    14/14 In the meantime, you can find our preprint here!

    Prikaži ovu nit
    Poništi
  12. 19. stu 2019.

    13/14 ... Confirming and generalizing these results will thus require (many) more studies.

    Prikaži ovu nit
    Poništi
  13. 19. stu 2019.

    12/14 A note on the limits of this study: i) it is based on scalp EEG: the underlying neuronal mechanisms thus remain unknown (e.g. adaptation, lateral inhibition etc); ii) we neither used naturalistic images nor did we vary presentation rate and subjects' task...

    Prikaži ovu nit
    Poništi
  14. 19. stu 2019.

    11/14 In sum, we demonstrate that ~4 successive images can be simultaneously decoded from brain activity. A particular hierarchical network specifically accounts for this computational property, and thus allows the brain to continuously make use of multiple snapshots of the past

    Prikaži ovu nit
    Poništi
  15. 19. stu 2019.

    10/14 We originally did not expect low-level features (i.e. stim orientation) to propagate in the hierarchy. However, post-hoc analyses showed that this phenomenon is also observed in deep conv networks (yes, we probably had bad expectations...)

    Prikaži ovu nit
    Poništi
  16. 19. stu 2019.

    9/14 Localizing each decoder confirmed that visual representations sequentially propagated across the multiple levels of the visual pathway

    Prikaži ovu nit
    Poništi
  17. 19. stu 2019.

    8/14 This architecture allows successive images to be simultaneously represented within distinct processing stages.

    Prikaži ovu nit
    Poništi
  18. 19. stu 2019.

    7/14 After simulating >1 billion models, only 2 architectures qualitatively explain the EEG dynamics of visual representations. Both suggest that visual information i) propagates in the network when the image changes and ii) is otherwise maintained by an activity-silent mechanism

    Prikaži ovu nit
    Poništi
  19. 19. stu 2019.

    6/14 Each architecture leads to specific dynamics, here characterized with Temporal Generalization (TG, see King 2014, , Myers and colleagues 2013 for more details)

    Prikaži ovu nit
    Poništi
  20. 19. stu 2019.

    5/14 A variety of neural architectures can explain this phenomena. We use a simple dynamical system modeling framework to tease them apart.

    Prikaži ovu nit
    Poništi
  21. 19. stu 2019.

    4/14 Question: how long are visual contents and visual flows represented in brain activity? Answer: ~1 sec, i.e. 4x longer than the stimulus duration. Surprising consequence: *successive* visual features can be *simultaneously* decoded at each time instant.

    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:

    ·