Super excited to announce my first ever published paper, because it has all my favorite stuff in it: ugly equations, clever ideas, large graphs, open source code... and industrial applications with a real impact!
Wanna know more? Buckle up! (1/9)
Our latest preprint is out 🎉 You like machine learning? You enjoy combinatorial optimization? You're gonna love machine learning with combinatorial optimization layers! (1/n)
Today I gave a tutorial about one of my favorite subjects, automatic differentiation. Cause everyone wants to do deep learning but no one wants to get down and dirty computing gradients. Unless maybe you do?
Conference abstract submitted two whole hours before the deadline! I'm a bit lost, what am I supposed to do with all this free time 😱?
Any advice
@AcademicChatter
@OpenAcademics
@PhDVoice
?
@tom_wilks
Awesome! I have a YouTube channel where I put song covers kinda like this one, would you want to collaborate and record it?
See for instance
Exciting career announcement: I was invited by Prof
@AlanEdelmanMIT
to visit his lab at
@MIT_CSAIL
this fall! I'm very grateful for this opportunity to work on the
@JuliaLanguage
ecosystem with some of the smartest people I know 😍
Just started a postdoc at
@EPFL_en
, where I will investigate
#ML
on
#graphs
. Many thanks to Patrick Thiran,
@KrzakalaF
and
@zdeborova
for having me here! Are any of you Twitter folks in the area?
Wanna learn how to program this stuff, and discover type systems as you do so? Check out the Pokémon-themed homework I made while teaching
@JuliaLanguage
at MIT!
give each pixel a random Pokemon type, and then battle pixels against their neighbors, updating each pixel with the winning type (using the Pokemon type chart)
we quickly see areas of fire > water > grass > fire, electric sweeping over, ground frontiers taking over etc etc
After tireless efforts by the reviewing committee, the program for
@JuliaConOrg
Local Eindhoven is finally online! The finest Julia talks, delivered right at your European doorstep 😍
Machine Learning folks: what are some typical applications where you would love to compute a Jacobian or Hessian (e.g. to solve linear systems) but it is too expensive in high dimension? Bonus points if the matrix is sparse!
Yes we are using
@JuliaLanguage
in
#MachineLearning
! The coolest application I know is putting funky layers inside your neural networks, typically our world-class solvers for differential equations () and optimization ().
Me: [sitting at a Paris cafe with Alan Edelman]
The guy at the next table: Sorry to bother you, I heard you two mention
@JuliaLanguage
, are you somehow involved in open source software?
Me: So, actually…
@icmlconf
be like: don’t worry about our climate impact. We made hundreds of researchers fly to Hawaii, but that’s okay because we (maybe) planted [checks notes] 36 trees.
Green machine learning research still has a loooong way to go.
Hidden Markov Models are a fundamental tool in statistics, whose applications go from bioinformatics to speech recognition. Sure, ChatGPT is fancy, but my latest work is some good old machine learning from the days of yore. Wanna know more? Dive in! (1/n)
Hello Twitter ! Pour donner un coup de pouce à mes camarades de
#doctorat
, j'ai commencé à rassembler ici quelques ressources utiles pour démarrer sereinement la
#recherche
🔬 Avis ? Suggestions ? Propositions ?
@BienDansMaThese
Here's to the one and only love of my life... my
#bibliography
manager
@zotero
📚 If you're a researcher and you don't use it yet, trust me: you will.
Thanks to
@ThibautRobine
for the violin track 💙
In case you missed it, here's the video of my
@JuliaConOrg
talk on implicit differentiation. TLDR: if you can dream it, you can differentiate it!
And of course, there's a
@JuliaLanguage
package too:
Exciting stuff happening in
#JuliaLang
world! A monthly newsletter has emerged thanks to
@StefanKrastanov
, keeping us updated on current projects and discussions 😍
It's like we always say: come for the speed, stay for the community (and also the speed).
Do you want to get started with automatic differentiation in Julia? If so,
@adr_hill
and I gave a beginner-friendly tutorial at
@JuliaConOrg
last week. Check it out and tell us what you think!
Me during my PhD: let's write some code to predict train failures from condition-monitoring data.
A random stranger today: btw I'm using your package to analyze swimming behavior in zebrafish.
Gotta love open source development.
Dear Santa,
I've been very nice this year. And for Christmas, I wish people would stop sharing links to Arxiv PDFs, and start sharing links to abstracts instead.
Thank you Santa.
Just in time for our introductory seminar at
@EPFL
!
What's different about this version of
#JuliaLang
? Simply put, it solves the number 1 gripe of every beginner: time to first plot ⏱️
I'm not saying Python should watch its back. But I'm not not saying it 😈
Dernière création musicale, dans la lignée de mes héros
@Les_Goguettes
. Quand la politique dépasse même
@le_gorafi
, il ne nous reste plus que la chanson.
Version complète sur YouTube... des idées pour une prochaine reprise ?
After months of procrastinating the final edits, I am pround to announce that my PhD dissertation is finally online: 🥳
If you're interested in
#MachineLearning
,
#OperationsResearch
or
#Railways
, feel free to take a look and reach out!
Thank you
@jure
for an inspiring keynote and for engaging with all the questions from the community!
Its exciting to hear about a new field being born
** Relational Deep Learning **
RelBench:
@wahajmb
@thephdstory
@PhDForum
@WriteThatPhD
@PhDVoice
I cannot stress enough how the choice of mentor and lab changes everything. Don't choose the subject, choose the workplace and the people. If you choose right, as I luckily did, then you're good to go. Starting my third year and I'm having a really good time despite the hardships
In Paris this week, giving two tutorials at the
#JuliaLang
and optimization days! Time to reunite with old nerd friends and make new ones 🤓 (where is the multiple dispatch emoji on this thing?)
@PhD_Genie
I’m a mathematics researcher, so everyone thinks I’m good at mental calculations. I guess their mental picture of math labs is a bunch of nerds in a cave challenging each other to perform long multiplications in record time 😅
Hey
#AcademicTwitter
! Currently preparing a math course for the fall semester, and I'm wondering how to make it more accessible to students with
#mentalhealth
or
#disability
issues. What is my margin of action as a TA? Any advice is welcome!
Ma dernière chanson parle d'un endroit merveilleux, plein d'humanité et de bienveillance, qui a contribué à la santé mentale et au bien-être de centaines d'étudiant·es en
#doctorat
. Venez !
#AcademicMentalHealth
#AcademicTwitter
To make things easier for you, we've developed a
@JuliaLanguage
package called InferOpt.jl, which turns any combinatorial optimization algorithm into a layer compatible with automatic differentiation.
You can try it out here: (3/n)
Starting my first full week without working from home since the pandemic began. My biggest worry: coping with the loss of my (AD-induced?) daily nap 😅 Friends from
#AcademicTwitter
, how did you manage the transition?
@OpenAcademics
@AcademicChatter
@PhDVoice
@stylewarning
People are just used to older languages, and if those work well enough there’s no incentive to change. For standard deep learning, Python remains your safest bet. For more general scientific computing,
@SciML_Org
and
@JuMPjl
will rock your world if you give them a try.
Heading to Berlin tomorrow to present our latest research on
#optimization
and
#deeplearning
. Excited to finally meet the names from my thesis bibliography! Plus I got a fancy talk title 😎
Thanks
@matbesancon
for the invite 🙏
French politics is too depressing right now, so let's talk about the pinnacle of computer science. Since 2007, the
@sigbovik
conference has proudly featured the most absurd, useless, extravagant and otherwise deranged papers in any scientific field I'm aware of. (1/n)
@ZJAyres
Hi there! Guillaume here, I study maths at a 🇫🇷 university and I'm the PhD student representative there. The Covid crisis and my own personal issues got me increasingly worried about the mental health of my peers and my students, so I try to help them however I can 😊
Un point de vue assez personnel sur la
#g
éologie🪨
Si vous aussi vous voulez que je me penche sur votre discipline de
#recherche
, suggérez-moi un titre dans les commentaires !
PS : on m'a conseillé de tagger
@Sciencecomptoir
, advienne que pourra...
@AcademicChatter
My go-to advice these days: disconnect and relax/sleep whenever you need to. You being exhausted or stressed helps no one, not even your work.
As for me, well, I wouldn't mind a more boring spring semester 😅
Guest keynote involving Boeing at
@JuliaConOrg
.
The speaker asks “What’s the one thing that drives better safety?”.
Someone from the audience confidently answers “Accidents!”.
@ZJAyres
Hey y'all!
I'm a math PhD student working on railway optimization, but the thing that brings me joy is music. I started writing (hopefully) funny song covers during lockdown, and it really helped me cope with all the stress. Here's a snippet of my latest project 🎶
My last song is for the kind of thesis advisors who think all it takes to mentor someone is to check in on them twice a year and sign their papers without reading them. Yes, you in the back, I see you 👀
#AcademicTwitter
#phdchat
To all the
@ProjectJupyter
notebook fanatics out there, ever heard about
@PlutoJL
? Maybe it's time for a change! The
#JuliaLang
community welcomes you with open arms... and dubious songs 🎶
Saw my psychiatrist today, and it seems I no longer need my trusted antidepressants to be healthy!
Of course my therapy will continue (as it should… for everyone?), but today I’m celebrating:
Guillaume 1 -
#OCD
0
Dernière création musicale, dans la lignée de mes héros
@Les_Goguettes
. Quand la politique dépasse même
@le_gorafi
, il ne nous reste plus que la chanson.
Version complète sur YouTube... des idées pour une prochaine reprise ?
@JuliaConOrg
in Boston: boring, mainstream, required a transatlantic flight 😴
JuliaCon Local in Eindhoven: exciting, hipstery, reachable by train from all over Europe 😍
#JuliaLang
friends, submit your talks before November 1st to make it a success! ⏳
It's official: my
#PhD
thesis has been approved by reviewers, and I'm going to defend it on Friday Dec. 16th at
@EcoledesPonts
🥳 Thanks
@YohannDCstr
and
@ParmentierAxel1
for getting me this far!
If you want to attend, either in person or remotely, shoot me a DM 📨
Bonne nouvelle du jour : j'ai reçu le prix de thèse maths-entreprises 2023 de l'AMIES ! Merci à
@YohannDCstr
et
@ParmentierAxel1
pour leur encadrement, et à
@GroupeSNCF
pour cette première collaboration, que j'espère poursuivre. Les maths du ferroviaire ont encore un bel avenir !
@OpenAcademics
Hi! My non-academic self is rather fond of:
🎸 rock music from the 70's and 80's
📺 funny TV shows, superhero movies
🎪 juggling and unicycling (sometimes both)
🧀 healthy food such as French cheese and chocolat cakes
🐦 doomscrolling
#AcademicTwitter
... see you there!
@OpenAcademics
Guillaume here, writing to you live from "La France" (aka the land of 🧀🥖🍷). I speak 🇫🇷, 🇩🇪 and 🇬🇧 (in chronological order), and I can blurt out some 🇪🇸 in life-or-death situations. I come from a beautiful volcanic region, for which I wrote this song
Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.
Career goal unlocked: I made the front page of the
@JuliaComputing
newsletter... with a song!
Guess there is more than one way to contribute to
#opensource
🤷
Julia Computing Newsletter January 2022: Get everything you need to know about Julia including the latest news, updates and blogs from across the globe.
#julialang
#Newsletter
Okay that's it, I'm declaring a six-month ban on new graph packages in the
@JuliaLanguage
. Enough is enough. I love the community but you people are goddamn exhausting 😂
Petit rappel avant les
#Manifs14aout
: le
#PasseSanitaire
, et plus spécifiquement le
#vaccin
, c'est notre seule option pour une rentrée à peu près normale. Je sais pas vous, mais moi j'ai bien envie de revoir mes élèves en amphi et mes collègues au bureau.
“There wasn't a point in the development where I thought "I can't do this". It just worked because Julia is a great tool for getting thoughts out of your head into the computer.”
New newsletter, about Julia! The gist:
- We should build more stuff in Julia
- Companies should use more Julia and hire more Julia devs
- You should learn to write Julia
Check it out here:
@OpenAcademics
Hi people 👋 Currently working on the Flatland challenge for my PhD! The goal is to manage railway traffic in real time and maximize punctuality... in spite of random train failures 🚂☠ Believe it or not, there are maths involved 🤯
Just received a negative answer for a postdoc application I submitted on December 16th, 2021.
That's it. That's the tweet.
#AcademicTwitter
#AcademicChatter
@ThePhDPlace
Choose your topic before your supervisor. And actually, don't choose your supervisor at all, just pick randomly. All supervisors are good and kind people, right?
Right?
Ever heard a pop song about complexity theory? Me neither, so I wrote one! Heavily inspired by my PhD advisor's favorite word 😉
To all the computer scientists and mathematicians out there: I take song requests 🎶 What would you like to hear next?