when a company advertises a "machine learning" or AI approach, we should ask them

- what is the model you're using? no, I won't accept that it's proprietary info.
- what is your training set? is the data complete? no, quantity is not good enough. where are the blind spots and how have you consulted with experts in the field to correct them?
- have you received peer review? If not, there's the door. come back when you have.


Previous post obtained more boost and favs that I could possibly hope for. Please do the same for my publications thanks.

(actually don't, smash the publish or perish mindset)


A recently released report from the ethical board of Axon (a company selling law-enforcement products) is highly critical on facial recognition software: static1.squarespace.com/static

An interesting reading to see how ethics are handled within a company.
I am happy with the board conclusions too, I just hope it will have more impact than just a consultative one.

Just finished the hackathon on constraint programming we started two days ago with my team at Constraint Programming Summer School! school.a4cp.org/summer2019/sch

We ranked 3rd, 2nd and 2nd on 13 teams. The goal was to propose a modelling and search algorithm to solve a almost real roadwork schedule problem. I really liked the hackathon format, which helped me to get onto the details real quick (I am a slow learner and terrible with theory)

Code if you want to check: github.com/GirardR1006/hackath

Yes, eclipse does not launch properly on my Ubuntu 18.04 release, it's time to do path hunting and jdk gathering. I love my job

Soon taking off to Wien for a one-week summer school on constraint programming 🛫

Hello fellows,

tomorrow, fine labmates of mine will present formal methods and their practical use during the french hacker festival "Pas Sage En Seine". Schedule is here (in french again): programme.passageenseine.fr/

They will be speaking about symbolic execution, automated design principles and other really cool stuff, for a non-scientific audience. If you happen to be in the neighborhood, just come here! There will be plenty of other cool talks as well.

Why do we want to verify our deep learning algorithm? To have guarantees that our programs actually perform the way we want them to do and to identify failure cases. On a wider scale, the aim is to better align our human goals and standards (economical, ethical) with the actual behaviour of our algorithms. Algorithms are human creations and can embbed all our biases, including unwanted (racist biases among others).
Being able to verify that our algorithm is not racist is actually a big topic

I totally forgot to mention, but I helped to organize a spring school bridging formal methods and deep learning. The aim was to provide an insight on how deep learning can help existing verification techniques, and how to adapt our verification tools for deep learning. We had some really cool presentations with talented speakers, and I hope it will spark some really interesting coopérations ;)

Video recordings and materials should be available soon here: formal-paris-saclay.fr/

Will probably begin to use Gnu Linear Programming Kit soon, any good resources to learn it over here?

#VimTips Or if you are already editing your .vimrc (which is likely since the only reason to source it is if you modified it since vim started), then you can do


because % stands for the current file.

Spent the afternoon trying to hunt for VRAM leaks in my torch code, failed miserably. What a cool day!

The crowdfunding of #Mobilizon has reached its first milestone in 5 days and that's really cool. I really hope it will reach the 2nd one for which Framasoft has pledged to make it federated through ActivityPub.

Nowadays instead of setting up a Doodle, most people I know set up a Framadate. Having a similar shift from Facebook events to Mobilizon would be great.


Show more
Scholar Social

Scholar Social is a microblogging platform for researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully. Read more ...