"they feel like such distant memories, like they were barely real"
Photo: @seckoraphotography
#cosplay #ccn #colossalconnorth #ccn2024 #colossalconnorth2024 #homestuck #jadeharley
I'm humbled - thank you Nico! I look forward to continuing the conversation we all started at #CCN2024.
(If Wikipedia is correct), Aschanffenburg should be very proud of its notable memory researchers.
"i would like to think that even if i was sad and scared, if i was put in a position where everyone depended on me, i could put all those feelings aside and do whats right!"
Photo: seckoraphotography
#cosplay #ccn #colossalconnorth #ccn2024 #colossalconnorth2024 #homestuck #jadeharley
@ernie
1 Liquid & Solid Brains
Mapping the Cognition Space
@ricard_sole
+
The debate over "understanding" in #AI #LLMs @melaniemitchell
https://mastodon.social/@teixi/113245780190962868
2 #DwarkeshPatel
w #DylanPatel & #JonY
https://mastodon.social/@teixi/113246369169408095
3 #CCN2024 […]
Not of this event. However, we did something overlapping at #CCN2024 - including the roundtable - and that was recorded. So many terrific comments there too!
Video link:
m.youtube.com/watch?v=m47q...
Abstract:
2024.ccneuro.org/keynote-nico...
Cosplays of alphinaud and Alisaie from final Fantasy 14 on a small carousel
Time to have fun after saving the world 😊
Alisaie: @faeriekirara.bsky.social
Alphinaud: Me ✨
📷: @lunya.bsky.social
[ #FF14 #FFXIV #ffxivcosplay #alphinaudleveilleur #alisaieleveilleur #cosplay #ccn2024]
Cosplay lineup for Colossalcon North Aerith Friday Honey Bee Inn Friday night Vi Saturday
Also I guess i can share a lineup since I'm literally packed lmao
#ccn2024
Reconstructions for seen and imagined stimuli.
fMRI vision reconstruction methods robustly generalize to mental imagery 2024.ccneuro.org/poster/?id=155 #CCN2024; PDF at 2024.ccneuro.org/pdf/308_Pape... "subjects imagined a small set of memorized target stimuli"
Shamelessly taking the opportunity to plug my own post and #CCN2024 talk! 😀bsky.app/profile/robmok.bsky.soci...
(btw, thanks for your #CCN2024 talk, it was great!)
Thanks @tyrellturing.bsky.social for not only posting on our paper but being 💯 on the mark in the replies. Big issue in finding & naming cells based on intuition & only publishing & studying those.
Do check my #CCN2024 @cogcompneuro.bsky.social talk! www.youtube.com/live/TS04Q-o...
#neuroscience
And that's a wrap on #CCN2024! I'd say it was a lovely, well-organized event highlighting a diversity of research areas. And I'd be remiss if I didn't mention the thing that many attendees raved about: a schedule on the back of the name tag!
Last talk! Using real time BOLD-based feedback in a navigation task to explore what is easy or hard to learn. Consistent with previous work, off manifold learning is hard! #CCN2024
Hippocampus encodes model-based choices during planning #CCN2024
Next we hear about the ability of mice to rapidly learn new maze environments. The first maze takes several trials to learn, but new mazes after that are learned much more quickly. Acortical mice take longer the first time but still generalize well & have long term memory! #CCN2024
Next talk tackles a perennial issue in neuroscience: we use simple cell types to describe neural function even though the story is more complicated than that. In ANN models, units with spatial responses weren't the most necessary for decoding spatial information #CCN2024
Final session of talks! First: On a collaborative two player game, a new version of a collaborative tree search algorithm best fits data. #CCN2024
Purely visually-trained models struggle to explain social processing, perhaps because they don't have the structure to represent how entities interact. What about using a graph neural network? #CCN2024
Is there a third hierarchical visual pathway for social scene processing? fMRI data suggests anterior regions are mainly modulated by social information #CCN2024
Now for Leyla Isik's keynote on social visual perception. Current models can struggle with understanding social interactions in images. What are they missing? #CCN2024
Noticing at #CCN2024 that machine learning people tend to use "learning"and neuro people tend to use "memory" when referring to similar things.
Changing theme for the final talk: how does sleep impact insight? The ability to figure out the rule in a challenging perceptual decision making task correlated with N2 sleep during a test period. Including aperiodic EEG activity can predict insight even better
#CCN2024
Data from humans and reinforcement learning agents playing a naturalistic navigation game shows that a model based on successor representations can capture human behavior and fMRI
#CCN2024
Talk 2: why does replay sometimes favor dispreferred options or distant memories? It could be due to differences in learning regimes, i.e. rich (small initial weights with slow learning) or lazy (large weights with fast learning) #CCN2024
Contributed talks! Value and certain go into decision making. What role does replay of last experience play in this? MEG shows replay does occur during deliberation, and is biased toward uncertain and aversive options (but this might actually drive irrational choices) #CCN2024
Open challenges: why aren't ANNs more impacted by curriculum learning? How should agents choose to learn? How are multi-purpose representations built? #CCN2024
Also he's running a summer school on these topics!
He is using this knowledge to understand processes like semantic learning in the brain. The learning of finer grain categories over time shows up as drops in the learning curves associated with different saddle nodes in the loss landscape #CCN2024