Emer Jones

@emercjones

Machine learning & cognitive neuroscience | PhD student at & ,

Vrijeme pridruživanja: siječanj 2010.

Tweetovi

Blokirali ste korisnika/cu @emercjones

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

  1. proslijedio/la je Tweet
    10. sij

    New preprint from the lab: "Individual differences among deep neural network models." Work with , , and Courtney Spoerer. below. 1/7

    Prikaži ovu nit
    Poništi
  2. proslijedio/la je Tweet
    8. lis 2019.

    Now out in PNAS! We combine MEG source reconstruction, RSA, and end-to-end deep neural network training to show that "Recurrence is required to capture the representational dynamics of the human visual system".

    Poništi
  3. 23. ruj 2019.

    Randi Starrfelt spoke about the Back of the Brain project: a large and rich dataset (64 patients, 32 tests). PCA stroke patients were recruited based on their lesions rather than their symptoms.

    Poništi
  4. 23. ruj 2019.

    Finally, an explanation for why there are cases of pure alexics and prosopagnosics if the processes are bilateral with dominant and not exclusive areas: people lie on a spectrum of lateralisation and those with extreme lateralisation are those in the dissociation case studies

    Prikaži ovu nit
    Poništi
  5. 23. ruj 2019.

    She proposed an explanation for why word- selective areas are more in the left hemisphere and face- selective regions are in the right hemisphere: 1. Local connections between word recognition and language areas (left); 2. face areas separate from word (become right-lateralised)

    Prikaži ovu nit
    Poništi
  6. 23. ruj 2019.

    Marlene discussed three common principles underlying face and word recognition: face and word selective areas are differentially, not exclusively tuned.

    Prikaži ovu nit
    Poništi
  7. 23. ruj 2019.

    Convolutional neural nets are a promising line of research regarding a more mechanistic account of the ventral stream, but don't yet speak to the specialisation, segregation and topography of the brain that we see empirically

    Prikaži ovu nit
    Poništi
  8. 23. ruj 2019.

    Marlene Behrmann opened the workshop with a great talk, discussing that the large "area" or "where" focus of central visual system research falls short of understanding HOW the brain accomplishes word and face recognition.

    Prikaži ovu nit
    Poništi
  9. 23. ruj 2019.

    Great programme for today's workshop on the visual ventral stream at the . Many thanks to and Randi Starrfelt for organising!

    Poništi
  10. 18. ruj 2019.

    Talk by on Deep Gaze 2's success with fixation prediction. Interesting that 4 model neurons account for most of DG2's accuracy (approx corresponding to object 'pop-out', text, faces, and 'geometry' respectively)

    Poništi
  11. 17. ruj 2019.

    Which network architecture is capable of capturing the dynamics of MEG data from different ventral stream regions? Recurrent architecture with bottom-up, top-down and lateral connections is necessary; ramping feedforward network is insufficient

    Poništi
  12. 17. ruj 2019.

    Talk by : Using UNet to segment timeseries data (saccade detection). Implemeted in PyTorch at

    Poništi
  13. 17. ruj 2019.

    New fMRI dataset for end-to-end modelling presented by . One subject, 30 episodes of Dr Who

    Poništi
  14. 17. ruj 2019.

    Great line-up of talks at the 'Deep learning in computational neuroscience' workshop at the !

    Poništi
  15. 13. ruj 2019.

    From 's tutorial on approximate inference on the brain, looking at whether we can explain human probability errors & biases in terms of different approximations of intractable Bayesian inference

    Prikaži ovu nit
    Poništi
  16. 13. ruj 2019.

    Fun fact: when asked for the last digit of their social security number, and then asked what year Gandhi was born in, the SSN digit influenced peoples' birth-year estimate (lower SSN digits -> earlier year). Crazy that a digit anchors a totally unrelated estimation.

    Prikaži ovu nit
    Poništi
  17. proslijedio/la je Tweet
    12. srp 2019.

    Help me out twitterverse, which team chat/lab organisational tool do you use/recommend and why? Slack? Basecamp? Mattermost? Telegram? (please RT for visibility)

    Prikaži ovu nit
    Poništi
  18. 6. srp 2019.

    Last post from ! Was a great week. ended the summer school with a very brief but super important reminder!

    Poništi
  19. 6. srp 2019.
    Prikaži ovu nit
    Poništi
  20. 6. srp 2019.

    Fab overview of deep learning-based image segmentation by incl U-Net, a top-performing model. Tips: pre-training is always beneficial, even if not domain-specific (ImageNet!) or the pretrained model doesn't initially match dimensions (eg greyscale/RGB)

    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:

    ·