tomdonoghue.bsky.social
Cognitive & Computational Neuro Scientist - studying electrophysiological signals in human brains, mostly by writing Python code. https://tomdonoghue.github.io/
41 posts
731 followers
298 following
Regular Contributor
Active Commenter
comment in response to
post
Oops - released**
comment in response to
post
I've been reading a lot of old papers and it's interesting to see how certain approaches dropped off, usually without any explicit "Away" papers - but seeing re-occurring patterns of things people try has made me wonder what a "history suggests this is a dead end" paper would look like...
comment in response to
post
That's why 1/f is foolproof! Flattened? Bigger high frequencies! Oh, it steepened? Bigger low frequencies! It's always better 😎
comment in response to
post
Short version: yeh, current evidence is not consistent with a simple "exponent == E/I" interpretation, it's more complicated
Longer version - I wrote a section on where I think E/I interpretations are at in this preprint (p.19 on "Interpretations"):
www.medrxiv.org/content/10.1...
comment in response to
post
An apple a day keeps the spectral tilt going the right way - that’s the saying right!?
comment in response to
post
Yeh, I screened that Gibbs & Gibbs paper for potential inclusion, but the criteria were explicit measurement / discussion of aperiodic / 1/f / broadband activity. Including papers that got close to doing / saying so (or could be explained as aperiodic) would be an intractable number
comment in response to
post
What is misleading about aperiodic?
The tool is not called one over f, it's now called 'specparam' for spectral parameterization. We've used the aperiodic & periodic terminology since the paper.
comment in response to
post
Basically I think 1/f brings a lot of baggage that isn't always helpful (from a physics sense, if one really wanted to evaluate this, it would imply it holds over broader ranges than tends to be the case). I take 'aperiodic' to be the descriptively appropriate but theoretically neutral option.
comment in response to
post
We've used the term 'aperiodic' since I've been writing on this, both because (as this paper notes) it's a longstanding term in the lit, but more because 1/f is a more specific term that is often somewhat wrong - most ephys data isn't strictly 1/f, it typically bends (is multifractal / Lorentzian)
comment in response to
post
Hopefully this little overview is useful in navigating the literature, and also surfacing a lot of knowledge sitting in this older papers!
Comments / suggestions welcome! Also do let me know if you think I've missed anything!
comment in response to
post
There are too many interesting papers to share here, but as well as the paper the project repository includes a chronological annotated bibliography, listing lots of interesting papers digging into aperiodic activity over the last 75 years:
github.com/TomDonoghue/...
comment in response to
post
In 1949, two separate papers reported that i) there was an exponential distribution of energy in the EEG (Motokawa) & that autocorrelation analyses suggested 'aperiodic motions' in EEG activity (Imahori & Suhara)!
In the 75 years since, there have been lots of cool papers, many seemingly forgotten.
comment in response to
post
That's awesome! I hope it's going well - feel free to reach out if you have any questions about the 2020 paper / tool, etc!
comment in response to
post
Also, I haven't really figured out the new 'starter pack' stuff - if anyone knows of any related starter packs in ephys stuff, let me know!
comment in response to
post
I'm very happy to have this project available - it's one I've worked on in bits & pieces for many years now!
Thoughts, comments & suggestions are welcome!
Big thanks to coathors Ryan Hammonds, Richard Gao, Leo Waschke, Eric Lybrand & Brad Voytek.
Link:
www.biorxiv.org/content/10.1...
comment in response to
post
Overall, we argue that across these methods & their relationships, there is a lot of information on 'aperiodic' activity (broadly construed), though it's currently disconnected. We hope that this methods exploration can help to connect between results and ideas in the literature.
comment in response to
post
This project uses basically all my tools & strategies - systematic literature searches (lisc), simulations (neurodsp), collecting various analysis methods, comparing to specparam (fooof) & empirical data analysis - all openly available in the project repo:
github.com/AperiodicMet...
comment in response to
post
The premise is that there are many different kinds of methods that measure (probably) related features of the data - especially when thinking of 'aperiodic' activity.
This project seeks to investigate how these different methods relate to each other!
comment in response to
post
Jokes aside, this is actually a cool paper arguing that then current practice was inadequate for demonstrating rhythmicity in time series since you can pull some random samples, process as standard, and "find" a rhythm. Good thing we solved that back in 1957!
comment in response to
post
From this paper:
link.springer.com/article/10.1...
comment in response to
post
This has been a wonderful collaboration between Mohamed's expertise in sleep & my work on spectral parameterization / aperiodic activity!
There's a lot in this paper (Mohamed did a ton of work!) so please check it out if it looks interesting!
Comments / questions / ideas welcome!
comment in response to
post
From this paper:
www.sciencedirect.com/science/arti...
comment in response to
post
Thanks for the recs! I went to Princeton Record Exchanges - what a cool place, like stepping back into a previous version of the music world hahaha
comment in response to
post
We developed the tool and have been using it mainly for human single-unit recordings - but the main components just generalize to animal models too!
Comments, questions, suggestion all welcome (preferably on Github!)
comment in response to
post
Eventually all that code did turn into a dissertation, but that was by no means the original plan lol
comment in response to
post
Ohhh, weird - thanks!