summerfieldlab

@summerfieldlab

Investigating the mechanisms that underpin human learning, perception and cognition, headed by Chris Summerfield

Oxford University
Vrijeme pridruživanja: siječanj 2013.

Medijski sadržaj

  1. 25. stu 2019.

    summerfieldlab joined the Oxford rally today in support of better working conditions in UK higher education. Here's Adam campaigning in support of Leurers (whatever they are)

  2. 16. stu 2019.
    Odgovor korisniku/ci

    this is a fascinating paper, thank you. So I'm >800 times more likely to obtain a significant result in my fMRI analysis if I use FSL rather than SPM? really???

  3. 15. lis 2019.

    solve one of machine learning's timeless problems: how to avoid being peturbed if you are attacked by a stuffed giraffe:

  4. 3. lis 2019.

    many congratulations to on a successful PhD viva!! and thanks to and Uta Noppeney for providing the hard questions.

  5. 22. ruj 2019.
    Odgovor korisniku/ci

    Agreed calculations can be complex. Here's one site I found helpful: . I'm going to Marseille on wednesday - here are the estimated emissions by rail, road & air

  6. 13. kol 2019.

    Congratulations to the lab’s newest Dr !

  7. 27. svi 2019.

    Tomorrow at Poster 68: Mira will show that distractors exert a multiplicative influence in perceptual decisions consistent with an adaptive gain account of feature encoding - come check it out!

  8. 16. tra 2019.

    Interestingly, simply assuming that more than one value dimension contributes to the agent’s distance-to-goal estimation (eq. 2) yields diminishing marginal utility and convex indifference curves over value dimensions. 10/13

    Prikaži ovu nit
  9. 16. tra 2019.

    The key intuition is that the agent represents an abstracted, multi-dimensional value map on which it defines its current position and goals. Thus, the agent will seek to minimise its distance to goal through interactions with the environment. 8/13

    Prikaži ovu nit
  10. 16. tra 2019.

    Distance to goal signals in a wider sense have recently been reported by us and many others across a range of regions in the medial prefrontal cortex, indicating that these signals are a ubiquitously tracked quantity in humans. 6/13

    Prikaži ovu nit
  11. 16. tra 2019.

    In this framework, rewards are “sensed” by the agent via dedicated input channels as if the environment furnished a ground-truth reward signal which the agent can interpret and maximize according to its preferences. 3/13

    Prikaži ovu nit
  12. 26. ruj 2018.

    Explaining AI to a curious dog (and some curious people too)

  13. 4. ruj 2018.

    We found that the BOLD signal in dACC, AIC and SPL were best explained by the context-modulated decision variable predicted by our model (compared to alternative models). This remains true after we partial out the influence of RT on BOLD signal. [21/n]

    Prikaži ovu nit
  14. 4. ruj 2018.

    Therefore, we ask what is the role of dACC and interconnected regions in adaptive gain control. We conducted an study using the more complex flanker task (where target and distractor strength and variance changed between each trial) [20/n]

    Prikaži ovu nit
  15. 4. ruj 2018.

    Our model counterintuitively predicts that under certain circumstances, there is a reversal of the conflict effect – that is, you are slower on congruent rather than incongruent trials (top-left corner).[18/n]

    Prikaži ovu nit
  16. 4. ruj 2018.

    Our model correctly predicts that the conflict effect disappears when the flankers are variable (Blue & Green lines). This effect is driven by congruent trials (‘cong’; when target and flankers agree). [15/n]

    Prikaži ovu nit
  17. 4. ruj 2018.

    One key feature of our model is that the degree of neural sharpening depends on info variability. Therefore, we devised a flanker task (pps respond to target tilt, ignoring the irrelevant flankers) where the surrounding flankers can be non-identical. [13/n]

    Prikaži ovu nit
  18. 4. ruj 2018.

    When participants are given less time to deliberate, they tend to make more “suboptimal” decisions – i.e. more influenced by the decoy. Our model is also able to account for this effect. 1st&3rd column reprinted from Pettibone (2012) and Trueblood et al. (2014) [12/n]

    Prikaži ovu nit
  19. 4. ruj 2018.

    It turns out participants do not always exhibit all three decoy effects and have stereotypical intercorrelation pattern on the decoy effects. Our model is able to replicate the effect. Panel A-D reprinted from Berkowitsch et al. (2014) [11/n]

    Prikaži ovu nit
  20. 4. ruj 2018.

    There are 3 types of decoy effects – compromise, attraction and similarity – which are determined by the attribute values of the decoys themselves (green dots). Our model accounts for all 3 effects. [10/n]

    Prikaži ovu nit

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:

    ·