DrYohanJohn

@DrYohanJohn@fediscience.org

Computational Neuroscientist

Reseach Assistant Professor at Boston University

I make biological neural network models of cognitive-emotional interaction in the limbic system. I'm also interested in philosophy of mind, emergence, and phenomenology.

Also, I sometimes dabble in mathematical concepts that are way over my head. :P

__

searchable

This profile is from a federated server and may be incomplete. Browse more on the original instance.

DrYohanJohn, to random

How seriously should we take "levels" of reality? I mean the parcellation of the universe into, say, the particle level, the chemical level, the cellular level, the organism, individual level, the social/ecological level etc?

Do levels interpenetrate?

Are these metaphysical matters or just pragmatic ones?

Stimulated by the great session on mental disorders just completed, hosted by @PessoaBrain , featuring Anneli Jefferson, @awaisaftab , @NicoleCRust , Alexey Tolchinky & @eikofried

DrYohanJohn,

@NicoleCRust

I'm glad you brought up the topics of causality! This is a place where a strict separation of levels becomes problematic from a pragmatic perspective.

We already have enough evidence to know that the social level, the subjective/experiential level, and the molecular level interact: alcohol clearly operates on all these levels.

So from a pragmatic perspective, causes belonging to one level can be accessed from other (usually higher) levels.

@PessoaBrain @awaisaftab @eikofried

DrYohanJohn,

@NicoleCRust @PessoaBrain @awaisaftab @eikofried

It is rarely the case that a problem can only be addressed via causes that belong to a particular level. It is true that talk-therapy can help with problems at the 'level' of mental self-talk, but sometimes medication works too.

Conversely, it may be that talking is the most precise 'scalpel' with which to manipulate certain receptor level events. :)

NicoleCRust, to random
@NicoleCRust@neuromatch.social avatar

How do you pack a 95K word (nonfiction) book into a 40 minute talk?

How many words are in 40 minutes? My estimate is 4-5K. That's ~20-fold compression. Something like half of 1 (of 10) chapters in the book.

Obviously you don't just read off the first half of the first chapter. But an outline of all of it is also super unsatisfying; it needs more depth than that. Clearly you present the central thesis and why it matters. But what to support it? This is a problem I've never encountered before. Not yet sure how to wrap my head around it.

Any advice? Any pointers to book talks you love?

DrYohanJohn,

@NicoleCRust

One of my favorite book talks:

https://www.youtube.com/watch?v=CZIINXhGDcs&pp=ygURZ3JhZWJlciBkZWJ0IGJvb2s%3D

Graeber highlights some big picture points from a book that's just full of fascinating details.

DrYohanJohn, to random

I can't be the only neuroscientist who thinks of this every time they see the word 'brainstem'.

Narf.

https://www.youtube.com/watch?v=snO68aJTOpM

DrYohanJohn, to random

Interesting paper linking phenomenology and psychophysics.

https://link.springer.com/article/10.1007/s11097-005-5852-6

๐Ÿงต

DrYohanJohn, to random

This seems cool, and relevant to our Goodhart paper.

"Here, we show that designing an agent in a modular
fashion as a collection of subagents, each dedicated to a separate need, powerfully enhanced the agentโ€™s capacity to satisfy its overall needs."

https://doi.org/10.1073/pnas.222118012

DrYohanJohn,

@seanpatrickphd

Whoops! Just fixed it.

DrYohanJohn, to Neuroscience

What do you think are the best critiques of the predictive processing framework in neuroscience?

I like the way that it has made top-down processing more popular in the minds of experimentalists. But I am skeptical that only errors are propagated up the hierarchies.

DrYohanJohn,

@NicoleCRust

Yup yup! I suppose you represent the spirit of biology, which, as I see it, embraces diversity and difference rather than attempt the spherical cow approach of physics.

The trick is to go the next step, dialectically, which Darwin did but few others do even today: turn the diversity into a key element in the explanation/model.

As a modeler I have to have a little sympathy for the spherical cow team though. :)

DrYohanJohn,

@NicoleCRust

Haha! Sure.

When a model is very simple, there is the possibility of working through all/most of its implications mathematically, and perhaps using it for coarse prediction. You can also test the logic of the hand-wavy bits of discussion sections. :P

Infinite series in calculus serve as an analogy: even the first two terms tell you a lot (when the phenomenon is repeatable). The model can become a lens with which you draw attention to the yet-to-be-explained data.

DrYohanJohn,

@skarthik @NicoleCRust

Exactly! Make the spherical cow and use it like a yardstick and/or lens.

DrYohanJohn, to Neuroscience

"However, when similar flight paths are compared across conditions, the stability of the hippocampal code persists."

Doesn't this mean that "place cell" is a misnomer (for such firing patterns)? They're really "place+trajectory" cells. Shouldn't a "true" place cell be path-invariant?

Credit-assigning on such invariant reps is presumably needed somewhere in the system in order to drive flexible navigation, no?

https://www.nature.com/articles/s41586-022-04560-0

DrYohanJohn,

@WorldImagining

Right but distance is already always a relation, either between landmarks or from a point to some "origin", which is also basically a landmark. There's no other sense of "place", really.

And a pure "distance from X" cell would have trajectory-invariance.

(The paper isn't on sci hub?)

DrYohanJohn,

@marcwhoward @WorldImagining

This make sense to me, but I also think that from a normative perspective it would be nice if there were path-invariant place cells (or a distributed with this property), even if they were strongly context variant... under the assumption that credit-assignment for place preference/avoidance should be at least somewhat path-invariant. Such representations could then guide navigation with that context, given contingent constraints.

DrYohanJohn,

@marcwhoward @WorldImagining

It also strikes me that very distributed representations might create problems for credit assignment: reinforcement would lead to a lot of feature-based overgeneralization.

In other words, it makes sense that conjunctive representations (logical AND like) exist and are relatively localized.

How might one design reinforcement on distributed reps that avoid/minimize such overgeneralization?

DrYohanJohn, to random

Very readable paper on the partially overlapping phenomenologies of meditation and psychedelic experience.

Milliรจre, Carhart-Harris, Roseman, Trautwein, and Berkovich-Ohana. "Psychedelics, meditation, and self-consciousness."

https://doi.org/10.3389/fpsyg.2018.01475

DrYohanJohn, to random

Most models that implement neural competition seem to use inhibition to achieve competition among neurons/ensembles.

What other plausible mechanisms for neural competition are there?

DrYohanJohn, to random

It seems very likely that similarities between real brains and artificial neural networks in performance and even feature-extraction have more to do with the shared data set and the presence of many free parameters than any structural or procedural commonality.

DrYohanJohn,

@NicoleCRust

I don't think the size of the training data matters, especially if it's clean data.

The point is this: equivalent performance can arise because of equivalent structure, but also because the external world has latent structure. ๐Ÿ™‚

I actually think this is the greatest long-term use of ANNs: even as black boxes they teach us about latent structure.

(I had no idea I was any sort of guru. ๐Ÿคฃ)

DrYohanJohn,

@neuralreckoning @tyrell_turing

Right! One might say that the epistemic gauntlet must be directly proportional to the number of parameters in the model! :)

DrYohanJohn, to Neuroscience

It was fun to be part of this 'representation collective'! Also looking forward to how the Transmitter develops. Quanta Magazine is so good.

'What are we talking about? Clarifying the fuzzy concept of representation in neuroscience and beyond'

https://www.thetransmitter.org/representation/what-are-we-talking-about-clarifying-the-fuzzy-concept-of-representation-in-neuroscience-and-beyond/

DrYohanJohn, to random

Apocryphal but still on-point decades later.

DrYohanJohn, to Neuroscience

Very readable paper on a somewhat neglected but theoretically fascinating topic: eligibility traces.

https://www.frontiersin.org/articles/10.3389/fncir.2018.00053/full

#Neuroscience

DrYohanJohn,

@tdverstynen

Yup yup! I hadn't revisited the data in a while: now there's a fair amount of indirect experimental evidence at the molecular level too. But not the "thing itself". :)

  • All
  • Subscribed
  • Moderated
  • Favorites
  • โ€ข
  • JUstTest
  • everett
  • magazineikmin
  • mdbf
  • thenastyranch
  • khanakhh
  • rosin
  • Youngstown
  • ethstaker
  • slotface
  • modclub
  • kavyap
  • DreamBathrooms
  • Durango
  • provamag3
  • ngwrru68w68
  • InstantRegret
  • tacticalgear
  • GTA5RPClips
  • cubers
  • normalnudes
  • osvaldo12
  • tester
  • anitta
  • cisconetworking
  • megavids
  • Leos
  • lostlight
  • All magazines
    •