Ilyes Khemakhem

machine learning phd alumni at @GatsbyUCL, @polytechnique alumni.

[Helena Cobban @helenacobban]

THREAD: around 10 days ago, @mouinrabbani commented that the #Houthis in #Yemen didn't even really need to *hit* any of the ships passing thru the Bab El Mandeb in order to have a huge effect on global trade...>

[Ilyes Khemakhem @ilkhem]

@ak_signe @EloiseHendy @lickedspoon @VICE You got it

[Jackson Hinkle 🇺🇸 @jacksonhinkle]

🇮🇱🇵🇸 Netanyahu-appointed Israeli Digital Spokesperson @HananyaNaftali just posted, then DELETED, a tweet admitting that Israel bombed the Baptist Hospital in Gaza, killing 500 civilians.

Only problem? Naftali reported that the IDF thought they hit a “terrorist base.”

Whoops.

[Ilyes Khemakhem @ilkhem]

@ccanonne_ Use vim + vimtex. Available on all platforms ;)

[Ilyes Khemakhem @ilkhem]

@Mira404 Zotero! Best reference manager

[Ilyes Khemakhem @ilkhem]

@WehenkelAntoine @glouppe @Aistats2020 Congrats Antoine! Great work!

[Ilyes Khemakhem @ilkhem]

@fdmatoz @ncclementi @shortstein You can clone your overleaf projects and work on them locally, then push when finished for your coauthors to see the change. This has the added bonus of versioning your latex projects, and keep a backup on the overleaf servers

[UCL CSML @uclcsml]

Our next talk will be given by Luigi Gresele @luigigres from @MPI_IS and Giancarlo Fissore from @Inria_Saclay tomorrow Friday 18 Dec at 2pm GMT on relative gradient optimization of the Jacobian term in unsupervised deep learning! https://t.co/Tpxh6C271T

[Ilyes Khemakhem @ilkhem]

Check out this excellent blog post by @noukoudashisoup on noise contrastive estimation!🔥

[noukou dashi soup @noukoudashisoup]

Started a blog with @MichaelArbel and @ilkhem. My first post is on noise contrastive estimation. https://t.co/2pBaIL1J3r

[Ilyes Khemakhem @ilkhem]

@VonHohenheimFMA Take care buddy!

[Ilyes Khemakhem @ilkhem]

Recent collab with Durk, @RicardoPMonti1 and Aapo 👇

[Durk Kingma @dpkingma]

New theory paper: https://t.co/IELEDDmjYe. We show rigorously that EBMs of the type E(x|y) = f(x)•g(y) are, under fairly mild conditions, (1) identifiable in functions f and g, and (2) universal conditional density approximators, generalizing previous forms of nonlinear ICA.

[Durk Kingma @dpkingma]

Our paper "Variational Autoencoders and Nonlinear ICA:
A Unifying Framework" has been accepted to AISTATS'20. With @ilkhem, Ricardo Pio Monti and Aapo Hyvarinen (UCL). Surprisingly strong and general identifiability results, with rigorous proofs! https://t.co/Y5rHnjCYzw

[Ilyes Khemakhem @ilkhem]

@hardmaru I agree! We've tried to apply this to our recent work (https://t.co/6Rm5b5lKeZ) and I'm satisfied with the result, especially because this is my first time attempting this exercise.

[Ilyes Khemakhem @ilkhem]

@michael_nielsen Do you keep track of all these amazing references in tools like Goodreads? If so can we have access to your public profile?

[Assert Arxiv @assertpub_]

"Variational Autoencoders and Nonlinear ICA: A Unifying Framework" is the #1 paper on Arxiv today in machine learning. Congrats @ilkhem. See it at -> https://t.co/IIaw849ZaN and https://t.co/Z4owDr003W. Please retweet.

[Daisuke Okanohara / 岡野原 大輔 @hillbig]

Nonlinear ICA is known as unidentifiable with data only. They show that VAE conditioned with additional observation (e.g., time index, class label) can identify true components up to trivial transformation, showing principled disentanglement ability of VAE https://t.co/8YO7PcoHzz

[Ilyes Khemakhem @ilkhem]

solid work 🔥👩‍🚒

[/MachineLearning @slashML]

Variational Autoencoders and Nonlinear ICA: A Unifying Framework https://t.co/eTgSuGwKZI

[Ilyes Khemakhem @ilkhem]

@mmaxinelucille Who needs a camera when you have a phone 🤷🏼‍♂️

[Ilyes Khemakhem @ilkhem]

@HoetreX @jujumk @knnnj_ @DeeCayy 😭🤡