Theoretical systems neuroscientist @UniofNottingham. Author of The Spike. Essays: https://t.co/mEOXZkpScW
Jun 28, 2020 • 6 tweets • 2 min read
The quote below sounds good, but that conclusion can’t be drawn from the linked paper.
Why? Because the paper assesses the frankly dodgy “magnitude based inference” approach to statistics used in sports science, and then makes this unwarranted leap at the end of its abstract /1
Well, it has to be calculated from a spreadsheet provided by the original authors. There are no formulae, and no statistical paper validating it has ever been published.
Jun 22, 2020 • 9 tweets • 6 min read
The rare upsides of 14 weeks of lockdown have been listening to lots of music & unearthing long forgotten artefacts in the house. My recent find combined both - and suggested I missed my calling as a compilation album sequencer.
Like music? Then this thread is for you...
2/ Long ago (well, 2007), I made 3 CDs for my soon-to-be-wife's long commute, each a tour of big tracks and hidden gems in rock and metal.
The only rule: they had to have a killer riff.
(And the other rule: I had to be able to play it).
So was born: "Incessant Riffing".
Dec 12, 2019 • 10 tweets • 5 min read
Latest pre-print from us (inc. @mathewe), on a simple question.
How do your inferences about neural activity, correlations, and coding depend on your choice of using the raw calcium time-series or deconvolving the spike-events from them?
A thread... biorxiv.org/content/10.110…2/ First challenge: find parameters for the deconvolution algorithms. We take 3 clever algorithms from people much smarter than I, and optimise parameter-fits for each to ground-truth data.
Encouragingly, all 3 can accurately recover known spike-trains, but...
Jul 30, 2019 • 4 tweets • 2 min read
Why Model The Brain?
Or: An Actual Scientific Case For Brain Simulations
On the ten year anniversary of Markram’s (in)famous TED talk, @edyong209 writes of the avalanche of criticism surrounding the project to build a simulation of the human brain
A short thread for those who want more details.... 1/
2/ @edyong209 dives into neuroscience’s strong rejection of this project, but doesn’t explain why.
One of the crucial technical criticisms is that they were building the wrong thing, as I outlined here: medium.com/the-spike/how-…
Feb 11, 2019 • 5 tweets • 2 min read
Very much enjoyed this project led by Tom Gilberston, who's a true curate's egg: a clinical neurologist who can code and fit reinforcement learning models to data.
We asked: Do dystonia patients have a learning deficit?
[thread] biorxiv.org/content/10.110…2/ Yes! Tom showed his dystonia patients had a selective reversal deficit:
They could learn an initial pair of stimulus-outcome mappings just fine.
But they were impaired at learning the reversed mappings, only when the outcome was a reward, not a loss!
Jan 17, 2019 • 7 tweets • 3 min read
We've a fun new pre-print on how to find out if there is "interesting" structure in your network
AKA finding out if and how your data network departs from your specified null model
Thread... arxiv.org/abs/1901.047471/ We use a spectral approach: we estimate the distribution of eigenvalues expected under the null model (using generative models), and compare that to the distribution in the actual data network
Data eigenvalues exceeding the predicted null model bounds = "interesting"
Dec 7, 2018 • 6 tweets • 2 min read
What can rats do without a cortex?
A lot!
For your reading pleasure, I've made a PDF of my tattered copy of Whishaw's classic book chapter "The Decorticate Rat" from 1990
While you download it from this link, some notes... drive.google.com/open?id=16kPnZ…
From the late 70s, Whishaw, Kolb and colleagues produced over 20 years worth of careful studies on the behavioural and learning capacity of rats without cortex.
Its an intimidating literature - and this 1990 book chapter, long out of print, is the only review I know about.
Aug 29, 2018 • 4 tweets • 1 min read
A short thread on stuff found in the brains of humans (because we are primates) and not in small lab rodents.
And so by the same logic must therefore also be "what makes us human"
Fantastic writers all, as this thread will show... medium.com/the-spike2/ Here's Ashley Juavinett (@analog_ashley) on the peculiar experience of being an experimental neuroscientist
New lab paper! Huge effort from @simaggi1 in a wonderful collaboration with @apeyrache using his data.
I call it "Learning to Remember" - of course I'm now going to explain why...
nature.com/articles/s4146…2/ Rats learned new rules in a Y-maze by trial and error. Unusually, recordings were made in the medial prefrontal cortex from the very first session. So we could ask lots of questions with these data about how PfC's roles in learning and memory intersect
Apr 18, 2018 • 4 tweets • 2 min read
There's a New Crazy at the ever febrile AI-neuroscience interface: Demiurge
A short thread on this brave new company
They posted this job ad. I checked the date three times: it wasn't posted April 1st
jobs.sciencecareers.org/job/478087/alg…
Their website reads like an AI Master's student's fever dream of a business plan for world domination by autocratic socialists
