Thomas F. Varley
banner
thosvarley.bsky.social
Thomas F. Varley
@thosvarley.bsky.social
Dual PhD: Complex Systems & Computational Neuroscience - Postdoc at UVM in the Vermont Complex Systems Institute.

Information theory, synergy, and emergence.

Connoisseur of collapse phenomena
Idk about Green's exact number, but I think it's a pretty huge insight that the algorithm for computing the poverty threshold is hilariously out-of-date for our current economy.
November 24, 2025 at 9:34 PM
Thanks to my PI @doctorjosh.bsky.social, and co-authors Pedro Mediano and Alice Patania for their expertise in information theory and algebraic topology respectively.
11/11
November 20, 2025 at 8:46 PM
I've been calling this "invisible" synergistic information "dark information" for a few years now. Stay tuned for a more comprehensive write-up!.
10/N
November 20, 2025 at 8:46 PM
This is consistent with prior work that I, and others, have done in this space, which suggests that many popular tools in neuroscience (functional connectivity networks, manifold learning, etc) are all systematically blind to synergistic information, only giving a partial view of the data. 9/N
November 20, 2025 at 8:46 PM
This is, imo, a blow to the "manifold hypothesis" popular in neuroscience. Brain data may be highly compressible, but that compression is going to lose important information represented in the joint-state of many variables. 8/N
November 20, 2025 at 8:46 PM
And finally, we found that manifold learning algorithms like PCA are only sensitive to higher-order redundancies, and lose synergistic information: it's easy to find sets of brain regions that *strongly* deviate from independence...but appear indistinguishable from noise under PCA. 7/N
November 20, 2025 at 8:46 PM
In fMRI data, ensembles that had more topologically rich structure (more cavities, longer-lived/larger cavities) had more synergy as well 6/N
November 20, 2025 at 8:46 PM
It turns out: yes, they are. Using a combination of synthetic data with known topologies (hollow spheres, toroids, etc), and naturalistic fMRI data, we find that synergistic information is strongly associated with higher-order topological structures.
5/N
November 20, 2025 at 8:46 PM
So when a topologist and an information theorist talk about "higher order structures"...are they talking about the same thing?
4/N
November 20, 2025 at 8:46 PM
Topological data analysis quantifies higher-order interactions in terms of structures in point clouds (cycles, voids, etc), while information theory considers how information is distributed over joint, marginal, and conditional distributions.
3/N
November 20, 2025 at 8:46 PM
"Higher-order" interactions are all the rage in complex systems right now - since they provide a formalism we can use to probe emergent phenomena.
But the two major mathematical frameworks descend from very different branches of the mathematical family tree!
2/N
November 20, 2025 at 8:46 PM
The limit isn't in cell signaling, but rather, diffusion. Xenobots have no vasculature, and so cells can only get oxygen, nutrients, etc from their environment via passive diffusion. If the bot is too large, the insides die.
November 12, 2025 at 12:41 PM
Knock on wood, but I think Pelosi II is going to get absolutely hosed in an embarrassing fashion. Her mother had decades of incumbency advantage - whereas this lady will just be a symbol of political aristocracy in a moment when populism is the order of the day.
November 11, 2025 at 1:59 PM
I wish this article mentioned that many of the patterns observed in psychedelic fMRI studies (e.g. increased Lempel-Ziv complexity, entropy etc) have also been observed in M/EEG studies. There's more going on in psychedelic science than just rsfMRI studies.
October 29, 2025 at 7:08 PM
My take away from this whole discourse is that Democrats don't know what they want (or can't agree on what they want). Do they want Trump But Blue? Do they want Woke With Degrees? Do they want Neoliberal Clinton-esque?
In contrast, Republican know exactly what they want: Trump + liberal tears.
October 29, 2025 at 3:44 PM
What does this mean for Hansen's "Acid test"? It seems like the post-El Nino cooling has been (at bets) moderate, and certainly not in the range he gives for past years (iirc the expected cooling would be something like 0.2-0.3C).
October 28, 2025 at 5:12 PM
This is great work...but it's worth pointing out that most of the really interesting findings (increased LZC/higher entropy, etc) have also been found in M/EEG and in vitro neural cultures. So it's good to be skeptical of BOLD, but there's been solid replication going on for a while.
October 28, 2025 at 12:16 PM
This citation policy is dumb as hell, but if *this* is really the straw that breaks Pinker et al's backs, I frankly have to question their integrity.
Apparently it's fine for rent-seeking publishers to make billions leeching on the work of others...but God forbid they get "woke"...
October 26, 2025 at 9:20 PM
"Admirable?" Spring Nature has reaped billions in revenue by extracting and pocketing unpaid expert labor from scientists who are locked into a system that requires their participation to maintain their careers.
October 26, 2025 at 9:20 PM