Day 44: Either Way, You’re Probably Fucked

Trump’s trade wars in cartograms 📈🗺️ https://t.co/SsAIrajhOE via @WSJ pic.twitter.com/X6owX24Hjk

— Yaryna Serkez (@iarynam) July 2, 2018

Quick voter file analysis of @Ocasio2018 's surprise win

1) Big age divide in white areas, but not non-white areas
2) Similar performance in white/hispanic/asian precincts, less well in black areas
3) Young white turnout was off the charts and that helped her *a lot* pic.twitter.com/WKCBfCO2uW

— ((David Shor)) (@davidshor) July 2, 2018

(for anyone who doesn't know about this) https://t.co/jsM7GMpDdV pic.twitter.com/iQL04SrT1t

— Joshua Loftus (@joftius) July 2, 2018

Rents in Tokyo are on par with the bustling metropolis of Asheville, NC. Rents in Osaka are lower than any major city in the United States. https://t.co/Wx2yEKI5Fv

— woolie (@woolie) July 2, 2018

original image and all credit to https://t.co/HLbi2wYH3z

— Geoff Boeing (@gboeing) July 2, 2018

Check out our #ICLR2018 paper on "natural" adversaries (ie from data manifold) for black-box classifiers. Crucial in #NLProc for grammatical and similar advs, eg for NLI and Google Translate! w/@_zhengli_ & @ddua17, pdf: https://t.co/E93iRELjmL, code: https://t.co/TXzsuGSqFz pic.twitter.com/Hm1ypMYGmZ

— Sameer Singh (@sameer_) April 30, 2018

"Co-Training of Audio and Video Representations from Self-Supervised Temporal Synchronization," Korbar et al.: https://t.co/XlOizuXwn2

Another clever idea for video self-supervision, like the colorization one from a few days ago

— Miles Brundage (@Miles_Brundage) July 3, 2018

Deep convolutional neural networks are great models of the visual system, but these static systems don't explain the temporal dynamics of real visual responses. So we built deep recurrent networks:@aran_nayebi @qbilius @SussilloDavid @NeuroAILab

Paper: https://t.co/QpheOUU7Dl pic.twitter.com/veyXeg5RGi

— Daniel Bear (@recursus) July 3, 2018

Nice summary! It seems I missed to check some papers... https://t.co/e0jAn6gYFk

— Jin-Hwa Kim (@jnhwkim) July 1, 2018

Researchers from @Stanford are using #deeplearning and @NVIDIA #GPUs to assess infrastructure quality in Africa from satellite imagery. https://t.co/3YFcv4HA0v pic.twitter.com/C7APXiIPh2

— NVIDIA AI Developer (@NVIDIAAIDev) July 2, 2018

ICYMI, 🌟 code-through (& animation 🎞)
"Recreating an Economist WDI Chart in R" ✏️ @PaulCampbell91https://t.co/fvK2aOdaXk #rstats #dataviz pic.twitter.com/XWQhbzuclO

— Mara Averick (@dataandme) July 2, 2018

👍 step-by-step & crazy cool map 🗺!
"Subplots in maps w/ ggplot2" 🖋 @ikashnitsky https://t.co/BfnM4JOhJp #rstats #maps #dataviz #ggplot2 pic.twitter.com/5aEvs8If0r

— Mara Averick (@dataandme) July 2, 2018

One of our most successful blog posts, by our #ML team leader @vadimlearning, shows the power of embeddings applied to source code identifiers.

Worth a read!#MLonCodehttps://t.co/tGGGWtz2AA

— source{d} (@sourcedtech) July 2, 2018

New post: "AdamW and Super-convergence is now the fastest way to train neural nets".

Practically important new results from the amazing @GuggerSylvain. Easily beats our winning DAWNBench approach!https://t.co/aAssLlKaf3

— Jeremy Howard (@jeremyphoward) July 2, 2018

Nice tutorial on #TensorFlow for those with a good understanding of #python and #ML https://t.co/OEYnBGteph

— Emilio Ferrara (@jabawack) July 2, 2018

bench, a new📦 for timing #rstats expressions

- High precision timers
- Tracks memory allocations, garbage collections
- Human readable units
- Nice defaults for plotting

On CRAN now!

Learn more at https://t.co/luAxaDSCAV and https://t.co/3Zzjs076vb pic.twitter.com/c764OxELc7

— Jim Hester (@jimhester_) July 2, 2018

Want a fascinating and richly structured public dataset to analyze? We just reinstated Kaggle's public data release, with automated daily updates! https://t.co/tPx9xeARyl pic.twitter.com/pds4BD0PMR

— Ben Hamner (@benhamner) June 28, 2018

Data alone isn’t ground truth. (Oh, and algorithms are not objective!) https://t.co/MnPpGM8dL2 https://t.co/YtWkLyVPcf

— Angela Bassa (@AngeBassa) July 2, 2018

New blog post published! 🚀 Today I interview Francois Chollet (@fchollet), creator of the #Keras deep learning library and Google AI researcher + engineer. Read the full interview here: https://t.co/D0ZkkgJXIc 👍 #DeepLearning #MachineLearning #Python #ArtificialIntelligence #AI pic.twitter.com/h40JSc036O

— Adrian Rosebrock (@PyImageSearch) July 2, 2018

"Ways to think about machine learning" --> that's an interesting analogy: "Excel didn't give us artificial accountants, Photoshop and Indesign didn’t give us artificial graphic designers and indeed steam engines didn’t give us artificial horses." https://t.co/lKIdDObzpf

— Sebastian Raschka (@rasbt) July 3, 2018

Here's a good list of neural net-generated video games (link for more).
I'd play Judgeboardies if I got to play RBG. https://t.co/e9YpGg0fvm pic.twitter.com/gwExtxWvZ2

— Janelle Shane (@JanelleCShane) July 1, 2018

Can someone please make a little game that just shows you incomplete NN training curves and you guess whether learning rate is small, high or just right? Besides these you would have the nuclear button like 'batchnorm' which clears the level whatever the answer. #askingforafriend

— Ferenc Huszár (@fhuszar) July 2, 2018

To a Bayesian, the universe is mostly certain that your existence is meaningless.

To a Frequentist, there are infinitely many universes and your existence is definitely meaningless in most of them.

Either way, you’re probably fucked.

— Nihilist Data Scientist (@nihilist_ds) July 2, 2018