@fabrice13@neuromatch.social
@fabrice13@neuromatch.social avatar

fabrice13

@fabrice13@neuromatch.social

I am a Neuroscience PhD student, coming from a Bioengineering BSc and Data Science MSc, and from a ML scientist job for a hi-tech quality control company.
As a junior scientist in the MedMaxLab@UniPD, I deal with multimodal healthcare data and ML (from stats to deep neural nets, self/un/supervised).
I code in Python, I long for Julia.

I am fascinated with Cybernetics, I am worried about global warming, oppression and capitalist economy. I boost and write about those too.

Loving many arts and sometimes making music, but more outside the fediverse.

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

fabrice13, to italy Italian
@fabrice13@neuromatch.social avatar

Sometimes I answer to job postings even though I'm a PhD student and researchers with contract. Most jobs are not compatible with my current duties, but I see it as both a chance to "stay in shape" and getting to know interesting and interested companies and people for future collaborations or employment.
Today I have to bid adieu to a company I've been keep an eye on for years.
I sent my CV and received a mail from HR, openong with my named misspelled, asking my current salary (which is predictable with high confidence by the position I've written on my CV) because «it is useful for us to know if your desired salary is in line with our budget».
I seriously do not think that's how it works and I'm telling that.
I may be petty about the misspelling (some may say I'm aware of microaggressions), but my desired salary ≠ my current salary and no salary budget information was disclosed to me in return.
Tech workers and Italian workers are particularly bad with salary negotiation, and that must change, this looked a bit past the line.

baldur, to random
@baldur@toot.cafe avatar

“How Good, Kind, Caring People Became The Bad Guys”

IMO the conclusion here is that most people who think they’re good, kind, and caring actually aren’t. https://www.okdoomer.io/thebadguys/

fabrice13,
@fabrice13@neuromatch.social avatar

@baldur I don't see how that conclusion would stem from the (very interesting) article, unless your comment is purposely ironic and I did not get it.
It's almost certainly true, at least because most people feel they're at least average good, kind, and caring, but the essay is stressing how we tend to project bad sentiments and intentions over good people denouncing bad things.
If you're sarcastic, sadly it is very close to the way more common cliché you are mocking and I didn't get it

jonny, to random
@jonny@neuromatch.social avatar

Nothing but frustration all day long but tomorrow it is my neighbors 100th birthday so I am excited to bring her flowers and have her tell me about her entire life again

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny when I lived in Milan I once met a tiny old lady on my way home, right at the exit of the supermarket, and she kindly but assertively asked me to hold her hand till her house (but no need to hold her bag because she could manage).
On the way a woman noticed us and told me amused "oh it's your turn today!".
Meanwhile the old lady told me her life story, she was in her nineties and she was one of the first women graduates, with a MSc in biology, and she could hold the conversationó about what I was supposed to do in bioengineering and after that.
Lovely and unusual, Milan is not reknown for slow and warm interactions among strangers, much less between old white ladies and young black gentlemen.
Hope she's in her 100s too now, but amazing life regardless

fabrice13, to random Italian
@fabrice13@neuromatch.social avatar

It's that time of the year again, where I don't know what tensors are multiplied in convolutional neural networks and how many free parameters are there...😮‍💨 to the pencil and sheet we go, to reinvent the wheel...🛞 🎡

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny it does but sometimes I do my little things off the 'puter first (I got on the 'puter later, verified I was right and I wanted to do something unusual indeed, and stayed up late because the 'puter is bad after dinner and also hypnagogic tensor product fantasies).
More on this later😈

computingnature, to bioinformatics

Despite all their hype in AI, we found that Transformers do not outperform for cellular segmentation tasks: https://www.biorxiv.org/content/10.1101/2024.04.06.587952v1

Cellpose outperforms winner of Neurips challenge (Ma et al 2024) on test set images. We show example images from the test set with Cellpose segmentations overlaid in yellow.

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny @computingnature and it's not only the domain thing, the Transformers hype has a sort of mystical aspect with the mathematical properties of the attention mechanism, it's been 4 years I think of constant press treating vision transformers as the only SOTA models, with ConvNets of comparable size and compute (if not lesser, i.e. better) achieving comparable results, anyway one can put it and look at it.
The fact we call the dot-product similarity "attention mechanism" has done wonders

lysander, to random

Sulla questione aborto riconosco di avere delle idee alquanto granitiche, però vorrei anche sentire altre campane perché probabilmente mi sta sfuggendo qualcosa.
Ho sentito mettere nel calderone del discorso "francia ha fatto bene e dovremmo farlo anche noi/ francia ha fatto male e non dovremmo farlo anche noi" anche il punto della crisi della natalità, nell'accezione di "se aumenta la facilità di ottenere un aborto si peggiora ulteriormente la crisi della natalità, e ciò è brutto".

Ora, io vorrei capire chi, e in quale universo può pensare che una frase del genere non tanto regga, perché ha senso, ma che sia umana.

È una frase che suggerisce che per il bene di qualcosa come boh, lo stato, la patria, è giusto che ci siano persone che -contro il loro desiderio- mettono al mondo altri esseri umani.
Magari trovandosi persino in una situazione economica/familiare/personale in cui questo costo è insostenibile.

Quindi, io mi trovo a rigettare qualunque tentativo di mettere diritto all'aborto e situazione di natalità nella stessa frase, per evidenti vizi di basilare umanità.
Però se ne discute, quindi un qualche senso deve averlo.
Cosa mi sta sfuggendo?

fabrice13,
@fabrice13@neuromatch.social avatar

@lysander penso che relativamente a questa risposta, l'unica cosa che ti sfugge (forse) è che gran parte dei natalisti non sono femministi o progressisti, purtroppo.
L'unico incentivo giusto alla natalità è l'incentivo alla natalità desiderata, quello protettivo verso la natalità anche non programmata ma comunque accettata dalla partoriente.
Chi fa quel discorso orribile ovviamente ha una visione incompleta e ristretta dell'aborto e dei motivi per farlo (salute, contesti immutabili e contesti mutevoli, si può abortire oggi e avere figli domani), e soprattutto di inferiorità delle donne.
Ed è un peccato in sé, ed è un peccato anche per chi, come me, pensa che l'antinatalismo sia "sbagliato" ma non passerebbe mai sopra ai diritti di chi esiste e lotta, solo per promuovere ideali proiettati su persone che ancora non esistono.
Forse pensano che due torti facciano una ragione, ma non stiamo parlando di dispetti, stiamo parlando della vita nel suo aspetto più intimo ed individuale e nel suo aspetto più collettivo ed esistenziale.

fabrice13, to random Italian
@fabrice13@neuromatch.social avatar

Of course being a black person on a train has always to come with an extra stress + double guessing.
Is it normal for white people to be disrespected and yelled at because you didn't download and screenshot the digital ticket's QR code and the train conductor has to wait 40seconds for the train to pass a region with no internet signal?
Mocking you like you have the arrogant presumption to have WiFi on board (which actually happens on a fair amount of trains even for Deutsche Bahn)?

jonny, to random
@jonny@neuromatch.social avatar

If you want to see a simple idea executed perfectly, or if you want to just get a lil hypnotized, here ya go
https://youtu.be/SthcxWPXG_E

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny saved for special meditations

fabrice13, to Blog Italian
@fabrice13@neuromatch.social avatar

What is the best way, platform or tool for an early stage researcher that wants a personal site to link repositories, show a cv, but also blog about papers and ideas informally?

fabrice13, to Cybersecurity Italian
@fabrice13@neuromatch.social avatar

What is a good password manager that I can use seamlessly on laptop and phone? Are there any free?
LastPass restricts the devices you can use it on, if you don't pay for premium. And had quite a few security problems...

fabrice13, to ArtificialIntelligence Italian
@fabrice13@neuromatch.social avatar

On vs and
Just skimmed through "Inferring neural activity before plasticity as a foundation for learning beyond backpropagation" by Yuhang Song et al. https://www.nature.com/articles/s41593-023-01514-1

Quite interesting but confusing, as I come from DL.
If I got it right, the authors focus on showing how and why biological neural networks would benefit from being Energy Based Models for Predictive Coding, instead of Feedforward Networks employing backpropagation.
I struggled to reach where they explain how to optimize a ConvNet in PyTorch as an EB model, but they do: there is an algorithm and formulae, but I'm curious about how long and stable training is, and whether all that generalizes to typical computer vision architectures (ResNets, MobileNets, ViTs, ...).
Code is also at https://github.com/YuhangSong/Prospective-Configuration

I would like to sit a few hours at my laptop and try to better see and understand, but I think in the next days I will go to Modern . These too are EB and there's an energy function that is optimised by the 's dot product attention.
I think I got what attention does in Transformers, so I'm quite curious to get in what sense it's equivalent to consolidating/retrieving patterns in a Dense Associative Memory. In general, I think we're treating memory wrong with our deep neural networks. I see most of them as sensory processing, shortcut to "reasoning" without short or long term memory surrogates, but I could see how some current features may serve similar purposes...

fabrice13, to Transformers Italian
@fabrice13@neuromatch.social avatar

The preprint of our lab's library for on data is out!
Check it at https://arxiv.org/abs/2401.05405

The repo (under review with the preprint for the amazing @joss ) is at https://github.com/MedMaxLab/selfEEG

If you want to try deep learning and EEG, if you have lots of data, but supervised learning is difficult or ineffective for your target task, you might want to experiment with self-supervised learning as popularized for and vision models!
Techniques such as MoCo, SimCLR are already implemented, and eeg augmentations can be used and further customized. If you do not know how to come up with architectures, don't worry! A model zoo is there 👨🏾‍💻🧠

fabrice13, to random Italian
@fabrice13@neuromatch.social avatar

Again, I'm having a hard time with and on Windows.

I wanted to follow along the juliaacademy course "Julia for Data Science", but the first tutorial shows how to activate an environment (?) but in the end I have no idea how to reliably work in an already installed JupyterLab, Julia and packages version compatible, and all in a specific environment for this project.
I've been installing and uninstalling Julia 1.6, 1.9 and 1.10, deleting them from PATH, etc.
With Python is not straightforward, but I got it.

I decided I will use and manually recreate all the courses notebooks. For the sake of learning, it's definitely better than just running readymade cells, plus it's in line with the MIT Julia course.

But it doesn't feel right, I feel there's a huge gap in my understanding, and likely a gap in the tutorials too.
Also, spending 2 days without proper coding, just running around bugs, is super frustrating, just feels like time lost forever to no cause.

fabrice13, to SciComm Italian
@fabrice13@neuromatch.social avatar

I remember one time I ...found a book w/ pdf.
It was about displaying dynamical systems theory, showing manifolds and so on.
Plots were not made with any programming language, they were actual drawings, pastels, watercolors.
It's the type of book I literally have dreams about, but I think this one exists😅 what was it?

Kindly boost my chances to find it 🙏🏾

fabrice13, to random Italian
@fabrice13@neuromatch.social avatar

During the holidays, i.e. from the 23rd to the 7th, I will not work.
I'd like to continue playing with the #julialang and some #pytorch but it's really for pleasure.
I will read some stuff about brains and perception but it'll be Ed Yong's "An Immense World", and perhaps McCulloch and Ashby.
Also, I'm finishing Dune.

If you too have the "privilege" of taking a break, what are you up to?

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny of course I do not know which battles and how you feel, but I wish you success in this, whether you pick them later on or not :)

In these months of starting my PhD and moving in with my gf, I needed to drop a battle I was picking and that I'll pick again, climate change activism; at the same time I dropped out of facebook and instagram, where the "battles" I picked where unfruitful and saddening discussions; and at work, with my family, friends and partner I had more big and small things I learnt not to pick, or how to pick...

fabrice13, to ArtificialIntelligence Italian
@fabrice13@neuromatch.social avatar

We're submitting a work to @joss, it is about #eeg and #selfsupervisedlearning #deeplearning

Are #neuro people interested in trying and using it? Maybe pointing to the best experts to review it even?
Let us know! 🧠👨🏾‍💻👩🏼‍💻🧑‍💻

#opensource #openscience

fabrice13, to Neuroscience Italian
@fabrice13@neuromatch.social avatar

Hello fellow scientists, I'm looking for academic journal suggestions!

Where would you submit, or where would you go read a paper about clustering subjects based on the similarity network built from their cognitive and motor tests scores after stroke?
Preprint at https://www.medrxiv.org/content/10.1101/2023.11.08.23297808v1
The paper discusses both methodology (similarity measures*, graph clustering**, compared to the tradition in the field of PCA and regression), and clinical aspects (typical lesions of cluster, diagnostic power of the assessment).
It is too computerish for some clinical neurology journal, not enough for some computers-in-med/bio/etc journal.
Must be

Ofc it's a team decision, won't be based on replies, but I am an early stage researcher and I really value discussions outside my lab!

*it's an old, little known one!
**it may be argued one of the techniques is new, both seem new wrt stroke

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny we were waiting publication/acceptance to publish the repo (but we might just open it soon independently, I'd like to also publish the data too but that's not on me...), I'll also check why the equation is unclear.
Anyway we checked and that distance measure is a semi-metric, exactly because the triangular inequality does not hold.
The discussion was left out of the paper, but we think violating the inequality is a good thing in this case, plus the clustering techniques both work without assuming the graph weights reflect distances in metric spaces.
d_{acj} is the sign function of the difference between subject a and subject c on feature j (when comparing subject a and subject b, you also compare both of them against any other subject c)

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny basically, we really liked the fact that the distance between two syndroms depends on whether there's a third one in-between, and we acknowledge we're not dealing with objects in a geometric space to start with.
The downside is that any pairwise proximity depends on the cohort and it's not meaningful/useful if working on a very small sample (but that's how things go with correlation measure)

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny oh, that's simply bad time management and mainly my fault, we wanted to have a nice clean repo with easy-to-use notebooks especially for medical doctors, and we tried a submission that just output the preprint so we let it out before code being as we wanted, then life happened.
I'd like so much to also publish data with it, which is another story... But it doesn't have to be all at once, so yeah, co-author and me are on it soon :)

jonny, to random
@jonny@neuromatch.social avatar

i am reading this document right now that i have no words for
https://www.kurims.kyoto-u.ac.jp/~motizuki/Essential%20Logical%20Structure%20of%20Inter-universal%20Teichmuller%20Theory.pdf

fabrice13,
@fabrice13@neuromatch.social avatar

@jonny I am almost moved to tears although I have no idea what it's about.
Reminds me of disputes and rebuttals I've seen somewhat followed e.g. with Schmidhuber vs Hinton, Bengio, LeCun, or Aaronson vs Tononi.
But I've no idea what is going on.

elduvelle, to productivity
@elduvelle@neuromatch.social avatar

I know this is not Linkedin, but I’m curious: what are the tools that help your personal the most?

[.] What works for me:

+] this basic pomodoro from @clive: http://text-pomodoro.glitch.me/ when I really need to focus

+] Zotero as a reference manager: https://www.zotero.org/, I would be lost without it

+] Github (desktop): my coding became much more focused, organized and reusable thanks to it: https://desktop.github.com/

+] GanttProject to realize how little time I have left… : https://www.ganttproject.biz/#is-easy

[.] What I tried but didn’t work for me:

-] Tasks “manager”: tried Obsidian, One Note, Google Keep, Notebook… I can’t seem to find The One that works for me (and syncs with my laptop & phone). Taking suggestions!

-] Something to read and annotate pdfs: I got a Remarkable2 tablet, and then an Onyx Boox Note Air 2, and they’re just too slow and do not easily sync with my laptop and Zotero. For the most important papers I have to say I might prefer paper.. Also taking suggestions for this.

Edit: trying to make the bullet points not disappear in is still hard

fabrice13,
@fabrice13@neuromatch.social avatar

@elduvelle @clive thanks for sharing (better here than on linkedin). I'll try the Gantt thing.
Not really sad to read Obsidian doesn't do it for you, but I'll note anyway that it is not a task manager although it has specific plugins for that.
As a note-taking and note-organizing app, I like it quite a lot, and use it quite a bit :) I'd like to be more intense with it (and note taking in general, btw), but while the developers community around it is wonderful, people explaining how to use it for productivity are mainly doing "look how used Obsidian to make this guide on how to use Obsidian".
To whom it may concern, I found her different, liked her stuff https://youtu.be/OUrOfIqvGS4?feature=shared

fabrice13, to Neuroscience Italian
@fabrice13@neuromatch.social avatar

1/n
Our pre-print is finally out!
Here's my first 🧵
In this work, co-authors and I clustered ischaemic stroke patients profiles, and recovered common patterns of cognitive, sensorimotor damage.

...Historically many focal lesions to specific cortical areas were associated with specific distinction, but most strokes involve subcortical regions and bring multivariate patterns of deficits.
To characterize those patterns, many studies have turned to correlation analysis, factor analysis, PCA, focusing on the relations among variables==domains of impairments...

https://www.medrxiv.org/content/10.1101/2023.11.08.23297808

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