←
Joris Falip94 posts
Joris Falip avatar

Joris Falip

@dataframefr

Associate professor working on machine learning at @UTTroyes (France). Loves longboarding 🛹, cyberpunk 🤖 and a lot of pop culture 🕹📚🎼 He/him

◷ Joined July 2009▧ 19 Photos and videos
Joris Falip avatar
Joris Falip@dataframefr·9 Jul 2019

High-dimensionality is a recurring nightmare of mine. But this morning at @UseR2019_Conf, we got a nice tutorial by @visnut on how to visualize high-dim data! Below is an historical depiction of me beating the curse of dimensionality, circa 2019 (colorized)

Joris Falip avatar
Joris Falip@dataframefr·8 Jul 2019

Tidyverse Dev Day is a wonderful moment to learn about #rstats 📦, thanks to everybody giving us a hand with how to get started contributing! @UseR2019_Conf really got everything, from 🦄 to 💻 and 🍺 ☕️ 🧁

Post attachment
Joris Falip avatar
Joris Falip@dataframefr·7 Jul 2019

On my way to @UseR2019_Conf for my first useR conference: excited is an understatement! With so many awesome #rstats talks, choosing which sessions to attend will be the hardest thing. Anyway, it’s adventuRe time! #chocolatine

Vince Buffalo avatar
⟳ Joris Falip retweeted
Vince Buffalo@vsbuffalo·4 Jul 2019

tbh, extracting knowledge from science papers remains a formidable challenge for most scientists too.

Extracting knowledge from science papers remains a formidable challenge for #AI. But V Tshitoyan @BerkeleyLab et al used unsupervised, deep neural net w/ embeddings instead of #NLP to show a promising path @cedergroup @Nature ★ @NatureNV by @olexandr: nature.com/articles/d41586-0…

Post attachmentPost attachmentPost attachment
Zachary Lipton avatar
⟳ Joris Falip retweeted
Zachary Lipton@zacharylipton·21 Jun 2019

I understand there's good reason to be careful w paper deadline times, e.g. not to subject researchers w. families to family time v. paper sprint crises. But for purely logistical matters, e.g., paper bidding, we should always adopt AoE time zone: one less thing to remember.

Joris Falip avatar
Joris Falip@dataframefr·19 Jun 2019

I finally took the time to read this paper and, like the "Mythos of Model Interpretability", it's one of those articles that challenge and redefine your perception of the subject! @zacharylipton is now definitely my spirit animal.

My talk on Troubling ***Trends in Machine Learning Scholarship*** today at IAS symposium on "Deep Learning: Alchemy or Science?" now up on YouTube: youtube.com/watch?v=A2Jtqi_o… #AlchemyOrScience [paper @ arxiv.org/abs/1807.03341]

Joris Falip avatar
Joris Falip@dataframefr·5 Jun 2019

And that’s not even including all the deeply entrenched biases and inequalities that we can’t currently get rid off. So setting them in stone by putting them in a decision system is probably the worst solution.

This tweet is unavailable