Nathanaël Perraudin (@NPerraudin) profile picture

Nathanaël Perraudin. Scientist

Joined August 2012
Photo by Nathanaël Perraudin (@NPerraudin)
A cool publication being the result of great teamwork at #sdsc Thank you to my co-authors!
Very happy to announce that our #paper “What You See is What You Classify: Black Box Attributions” was accepted at #NeurIPS2022. We tackle a fundamental problem in modern #AI & #ML: a lack of transparency of black-box models. Here: datascience.ch/black-box-att… @fperezcruz @OsaeKomi
4
Do you wonder how the recent heavy rains and thunderstorms affect Swiss lakes? Checkout our blog post bit.ly/3yt8hRK to learn about how we accurately simulate and forecast lake dynamics, & see our user-friendly portal!@EawagResearch #DataScience #Bayesian #3DSimulation
4
5
[Advanced] Intuitively, one could think that the projection onto the unit simplex is a simple rescaling. In most of the cases, it is not! You can check algorithms for the projection here. arxiv.org/pdf/1101.6081.pdf Old python2 code: gist.github.com/mblondel/6f3…
2
Nathanaël Perraudin retweeted
Conditional independence of entries of Gaussian vectors defines a graph associated to the inverse of the covariance (precision matrix). Graphical Lasso estimates a sparse graph by l^1 penalizing the maximum likelihood estimator. en.wikipedia.org/wiki/Graphi…
4
41
213
I love how problems were solved before the computer era. Such an elegant solution! Thanks @gabrielpeyre for the Oldies but goldies series...
Oldies but goldies: Joukowsky, N.E., Über die Konturen der Tragflächen der Drachenflieger, 1910. The Joukowsky conformal map solves airfoil design in the pre-computer era. en.wikipedia.org/wiki/Joukow…
1
[Advanced] Curious how the projected gradient descent method can be derived? Nice notes by Thibaut Lienart. tlienart.github.io/pub/csml/…
4
Definitely a very useful tool. But how much will our politicians consider it when taking decisions?
[Advanced] Want to learn about Sinkhorn divergences? Nice presentation by Aude Genevay @a_gnv: youtube.com/watch?v=TFdIJib_… Paper: proceedings.mlr.press/v84/ge…
Nathanaël Perraudin retweeted
Three weeks left to apply! Looking forward to extend our team and design proteins with AI.
We are #hiring a postdoc🧐to work on the intersection between #AI and computational protein design. The project will be carried out at EPFL in collaboration with @befcorreia, @mmbronstein, @trekkinglemon, and @SDSCdatascience. More information below 👇 Spread the word!🗣️
2
4
1
13
Nathanaël Perraudin retweeted
▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 100%
361
19,477
2,491
133,105
Nathanaël Perraudin retweeted
You asked for an efficient algorithm for non-convex optimization for Christmas? It won’t be possible unless you have a lot of smoothness. See why in this month blog post. francisbach.com/optimization…
5
90
4
555
Nathanaël Perraudin retweeted
The spherical Laplacian defines a set of orthogonal eigenspaces spanned by spherical harmonics. Display here is a random evolution in each space. en.wikipedia.org/wiki/Spheri…
2
75
7
393
Nathanaël Perraudin retweeted
How roots influence #ClimateChange: Soils play a major role for a stable state of the #Earth system. Milena Perraudin investigated in her master's thesis one of the most hidden parts of the terrestrial carbon cycle. geo.uzh.ch/en/department/125… #125GIUZ #soilcarbon
1
3
2
Nathanaël Perraudin retweeted
Principal component analysis and non-negative matrix factorization are two particular of non-convex matrix factorization problem. They differ by the constraints on the atoms D and coefficients X. en.wikipedia.org/wiki/Princi… en.wikipedia.org/wiki/Non-ne…
3
147
4
644
When I'm sometimes asked 'When will there be enough [women on the Supreme Court]?' and I say 'When there are nine,' people are shocked. But there'd been nine men, and nobody's ever raised a question about that." RIP #RBG🙏 pbs.org/newshour/show/justic…
2
9
1
63