Day 38: Tufte Controversy

For the many of us who are blocked by ET pic.twitter.com/blFC30zh3C

— ᴘʜɪʟɪᴘ sʜᴇᴍᴇʟʟᴀ (@philshem) June 26, 2018

Read our interview with Kaggle's first ever Kernels Grandmaster, Martin Henze (AKA Heads or Tails). He shares his story and practical tips for writing compelling EDAs. https://t.co/sSJpFNxo8c

— Kaggle (@kaggle) June 26, 2018

Welcome back, gradients! This method is orders of magnitude faster than state-of-the-art non-differentiable techniques.

DARTS: Differentiable Architecture Search by Hanxiao Liu, Karen Simonyan, and Yiming Yang.

Paper: https://t.co/gnKLXx6Pi9
Code: https://t.co/fZYIYNhLzz pic.twitter.com/pIHg3krnAE

— Oriol Vinyals (@OriolVinyalsML) June 26, 2018

"remarkable architecture search efficiency (with 4 GPUs: 2.83% error on CIFAR10 in 1 day; 56.1 perplexity on PTB in 6 hours)"
Try it now from: https://t.co/8khIix99mahttps://t.co/HFuW0II5Hl

— PyTorch (@PyTorch) June 26, 2018

This looks quite encouraging. Still some room for improvement in the results, but a good direction for Neural Architecture Search https://t.co/fJbkOMptAr

— Jeremy Howard (@jeremyphoward) June 26, 2018

Check out the code for our implementation of "Importance Weighted Actor-Learner Architectures", open sourced today!

Code: https://t.co/Uy56QCFG6X
Original paper: https://t.co/0nLsNV6wek

— DeepMind (@DeepMindAI) June 26, 2018

Tracking Emerges by Colorizing Videos. Vondrick et al presented this at the GAN tutorial, and you’ve got to see the videos to believe it! https://t.co/WovNHz3F6v #computervision pic.twitter.com/6rCdwGsRhT

— Tomasz Malisiewicz (@quantombone) June 26, 2018

Our latest work shows that learning to colorize videos causes visual tracking to emerge automatically!

Blog: https://t.co/FDVzJmmZ7h
Paper: https://t.co/U4jS83iI7B@alirezafathi @kevskibombom @sguada @abhi2610 pic.twitter.com/R3vMR3raFJ

— Carl Vondrick (@cvondrick) June 27, 2018

A research opportunity: Cyclical layer learning rates.
Follow along at https://t.co/adUt5ngnLf

— Leslie Smith (@lnsmith613) June 26, 2018

Our paper "Deep Convolutional Neural Networks for Breast Cancer Histology Image Analysis" was published in #ICIAR2018 conference proceedings
Text: https://t.co/GUn5rdD22T
Source code: https://t.co/fzqE4blkCL
joint work w/ @ARakhlin @shvetsiya @viglovikov#deeplearning #histology pic.twitter.com/30miInKmnj

— Alexandr Kalinin (@alxndrkalinin) June 26, 2018

While the number of Democratic women in Congress continually increases, the number of Republican women hasn't increased appreciably in almost 20 years. https://t.co/WAYlEBGL7R pic.twitter.com/Jd7jTFFU0I

— Nate Silver (@NateSilver538) June 25, 2018

Kennedy is currently the only true moderate on a court that has seen widening ideological polarization over the past few years. https://t.co/jhtPaGGRN6 pic.twitter.com/DLVkYE98xN

— FiveThirtyEight (@FiveThirtyEight) June 26, 2018

Global military expenditure as a share of global GDP:
1960: 6%
2016: 2.2%

From @eortizospina’s "Long-run trends in military spending and personnel: four key facts from new data” where you find the statistics for all countrieshttps://t.co/45xNJEsX6M pic.twitter.com/zXsoG0wTMM

— Max Roser (@MaxCRoser) June 26, 2018

Planetary orbits for the Heliocentric and Geocentric systems visualized. 🤔 pic.twitter.com/ZWifPrSzmr

— Fermat's Library (@fermatslibrary) May 23, 2018

Wow, decapitations in London have really gone up! #TimeOut #LDN #DataViz pic.twitter.com/DRDqCUTEFV

— Mike Brondbjerg (@mikebrondbjerg) June 26, 2018

My attempt to unify color palette usage in #rstats: paletteer! Access over 650 palettes from 27 packages using a simple interface 📦 https://t.co/XKUNqwdvOZ pic.twitter.com/60bV6czauK

— Emil Hvitfeldt (@Emil_Hvitfeldt) June 27, 2018

@thosjleeper's slopegraph package is pretty helpful…https://t.co/KsP9gfmdL4

— Mara Averick (@dataandme) June 26, 2018

So many counterexamples to this ignorance! Recently: https://t.co/3EAPV7S8Jx by @kjhealy and https://t.co/CaPks4XAlp by @ClausWilke. For >10 years: https://t.co/FQjWhVnZcK by @spatialanalysis. And much much much more! https://t.co/zRmhv5Rwzh

— Hadley Wickham (@hadleywickham) June 26, 2018

We've been creating graphics purely in R at @BBCNews for some time now - in consultation with our design colleagues and they are def not clunky!. Eg this from @nassos_ https://t.co/vXR6F4lPvY … and this from @cguibourg https://t.co/UL2p63dctN … & many more #rstats #ddj #dataviz https://t.co/yGOBHkmIJ5

— Christine Jeavans (@chrisjeavans) June 26, 2018

Man, it's a shame @EdwardTufte said this is impossible to do in #rstats with just code. I guess I shld delete my blog post. https://t.co/sw5CHXxViY pic.twitter.com/QC2FlgP1oT

— hrbrmstr (@hrbrmstr) June 26, 2018

After Edward Tufte's remarks about graphs and text in R, by happy coincidence I just got the proofs for my @PrincetonUPress book. All the graphs were made programmatically. The pipeline for the book ms and the website (https://t.co/NNfOGc4f5l ) start from the same Rmd sources. pic.twitter.com/UQSbmDQHC5

— Kieran Healy (@kjhealy) June 27, 2018

ICYMI, Charts à la @EdwardTufte in base, lattice & ggplot2 w/ code:
"Tufte in R" by @lukaszpiwek https://t.co/wfcSNAPsL2 #rstats #dataviz #infovis pic.twitter.com/X6Us6m3iSd

— Mara Averick (@dataandme) June 26, 2018

Five Principles for Programming Languages for Learners https://t.co/3mkaC6gUac

— Mark Guzdial (@guzdial) June 14, 2016

I made a dendrogram from chapters 2 + 3 of @tamaramunzner's 'Visualization Analysis and Design,' showing the what/why/how of #dataviz. The book goes deep, but I find this tree outline really helpful when thinking through + making new viz. https://t.co/b5uuJaQamd pic.twitter.com/S2y7i1dM71

— Jill Hubley (@Jill_hubley) June 26, 2018

I made a YT playlist on #rstats package development: https://t.co/7Crx2LFWt7

Includes unit testing, continuous integration, the usethis, covr, and devtools packages from @rstudio and others.

— John Muschelli (@StrictlyStat) June 25, 2018

This is good content (thanks!), but - if you're learning @TensorFlow today - I recommend skipping the graph level stuff - and beginning with tf.keras and eager - unless you have a specific reason to use this older style.

* https://t.co/mf4eZxngxi
* https://t.co/XkiVgWczBv https://t.co/gACHuRrFNn

— Josh Gordon (@random_forests) June 26, 2018

Three patterns for fast prototyping and research in #TensorFlow! https://t.co/78onjX5utv

— Danijar Hafner (@danijarh) June 26, 2018

Ah yes, the pinnacle of machine learning: the single-class classifier. https://t.co/Y51mQei60x

— Emily G kmii (@EmilyGorcenski) June 26, 2018

This isn’t algorithmic bias, like many systems it’s the hiding of human opinion and bias behind the veneer of an algorithm. https://t.co/yzqYqOHkkB

— Peter Skomoroch (@peteskomoroch) June 26, 2018

Here’s my previous thread on weaponized algorithmic bias and the systemic injustices we need to recognize before we can address it: https://t.co/Z1KLGy5SKI

— 🏳️‍🌈 Janus Cassandra 🏴 (@zenalbatross) June 26, 2018

Most #DataScience folks eventually learn that while we all aspire to create #dataviz that meets Tufte’s approval, we are all doomed to fall short under his watchful gaze.

So as with all other things in life it’s healthier to give up. Embrace your failure, for it is inevitable.

— Nihilist Data Scientist (@nihilist_ds) June 26, 2018

This guy runs a facial-recognition startup. His software doesn't work on his own face, so he demos it using a blond colleague. This is a problem https://t.co/YdVQDnrogu pic.twitter.com/fjGxeQB51Z

— Mark Milian (@markmilian) June 26, 2018

In 2017, there was a contest called "Learning to Run" at NIPS that produced gems like this: https://t.co/8D7BBnKnrr

In 2018, it returns, but now with a prosthetic foot: https://t.co/F8XXQE4uTM

objectives include "Deep Reinforcement Learning to solve problems in medicine"

— Sergey Levine (@svlevine) June 26, 2018

When models learn to “collaborate and communicate” it looks impressive. However, that’s only because these things are difficult for us because of human nature. They aren’t actually difficult to optimize for algorithms.

— Denny Britz (@dennybritz) June 27, 2018

Together with @goodfellow_ian and Patrick McDaniel, we wrote a CACM article on making ML robust against adversarial inputs. It highlights the need for more verification to complement current testing practices (e.g., benchmarking w/ CleverHans). It's here: https://t.co/2dYSBkAe2l https://t.co/1eXcnOTOn9

— Nicolas Papernot (@NicolasPapernot) June 26, 2018

Marking the end of my blog series on the topic (https://t.co/wPpx2ookdd), here’s my new survey of Actionable Intelligence: RL, continuous control, and their interplay: https://t.co/GU04JQNxsh

— Ben Recht (@beenwrekt) June 26, 2018

This is #MachineLearning / evolutionary computation in a nutshell. pic.twitter.com/lX2gpk8H98

— Randy Olson (@randal_olson) March 15, 2017

Thread with one path to becoming a data scientist.

It's so cool to see the variety of ways people enter this field. https://t.co/n4uYw5yvF1

— Data Science Renee (@BecomingDataSci) June 26, 2018

I wrote about ✨machine learning✨ in the 2018 @StackOverflow Developer Survey for @jaxentercom https://t.co/VMO3DsOLXJ pic.twitter.com/7R2JJVgrIp

— Julia Silge (@juliasilge) June 26, 2018

Blog post: "How open is too open?" https://t.co/EMcicxt6eJ - on open source projects, and sustainability.

— Titus Brown (@ctitusbrown) June 26, 2018

AI bias is one of our industry's greatest challenges. We have to build AI systems that hear all voices and recognize all faces equally across our diverse world to create the best future for everyone. https://t.co/ZL7C3Ik3v8

— harryshum (@harryshum) June 26, 2018

Post Edited: What is the role of statistics in a machine-learning world? https://t.co/oBq7y0ohWn

— Andrew Gelman (@StatModeling) June 26, 2018