The French election last Sunday has led to a few interesting data viz and data analysis projects such as the Economics chart on Where Emmanuel Macron’s Support Came From, an analysis of the numbers in Paris by Spreadsheet Journalism, or detailed maps by the New York Times showing How France Voted.
More generally though, Five Thirty Eight was critical of the role of data in these elections, arguing that it represented a failure for Polls, mostly because conventional reasoning had contaminated hard numbers.
If you’re more interested in American Politics, you can read the detailed data-driven article by Politico on how the media bubble is worse that people think, or appreciate this tutorial for mapping House & Senate votes in R.
And if you want to read about non-vote related data articles, read on!
Thinking material: Systems of Intelligence: Is this the VC meta-thesis we’ve been looking for?
ML for padawans: Sorting out the Basics Behind Sorting Algorithms
New! Facebook Research: A novel approach to neural machine translation
Data viz principles: Why You Don’t Believe In Facts, And How To Fix It
Here’s an AI goodie: Google Research has led to a feature that turns your selfies into personalized stickers
To be very honest, this week we found LOTS of great links that should have made it front and center in our newsletter. Here are those we left aside, but really wish we hadn’t.
- Automated Machine Learning — A Paradigm Shift That Accelerates Data Scientist Productivity @ Airbnb - by Airbnb Engineering & Data Science
- Your tl;dr by an ai: a deep reinforced model for abstractive summarization - by Salesforce Metamind
- Using Deep Learning at Scale in Twitter’s Timelines - by Twitter
- Simple is hard - by Mapzen
- [Paper] Gender differences and bias in open source: pull request acceptance of women versus men - by PeerJ
- General Tips for Web Scraping with Python - by Big-Ish Data
- Managing Spark and Kafka Pipelines - by Larry Murdock
- Magic AI: These are the optical illusions that trick, fool, and flummox computers - by the Verge
- How Privacy Became a Commodity for the Rich and Powerful - by the New York Times Magazine
- Generating Music with AI - by Lawson He
This is a recap of the links we shared in our weekly data science newsletter. Sign-up now to get it fresh, next Friday.
We’re all pretty aware that machine learning algorithms and data visualizations are unfortunately picking up human biases. But we may not realise how much yet.
The New Inquiry has released a paper and online map that predicts White Collar Crime, similar to all of those maps you keep seeing predicting crime. Except in this case, instead of criminalising poverty, they criminalise wealth.
Going further than being aware of our biases, we also have to be aware of the biases we have about being biased. Indeed, cognitive dissonance and other biases can mess with us even if we’re aware of it.
Could the answer to this issue be to turn towards unbiasedly algorithms to look at biased datasets, unbiasedly?
These problems are ever more pressing in a world where data is the new oil, and regulations have not yet evolved accordingly.
Also, here are some of the many links that didn’t make it into our newsletter this week, but are still well worth a read.
- 35 Years Of American Death by FiveThirtyEight
- What Government Corruption Scores Can Teach Us about the World by Vault analytics
- Leaked documents reveal how the government will demand your data under the Snooper’s Charter by Wired
- An AI just defeated human fighter pilots in an air combat simulator byBinary Loom
- Data preparation in the age of deep learning by o'Reilly
- Examining the arc of 100,000 stories: a tidy analysis by Variance Explained
- This neural network could make animations in games a little less awkward by Techcrunch
- How to Use Weight Regularization with LSTM Networks for Time Series Forecasting by Machine Learning Mastery
- How to Read Mathematics by Stonehill
- Same Stats, Different Graphs: Generating Datasets with Varied Appearance and Identical Statistics through Simulated Annealing byAutodesk Research
- Why we study interpretable units by Net Dissect
- #Awesome A Voice Kit by Google
RESPECT - Calling Bullshit - In the Age of Big Data
Cry out against Deep Learning - An Adversarial Review of “Adversarial Generation of Natural Language”
Protecting Privacy - Safe Crime Prediction
Dating Insights - Undressed: What’s the Deal With the Age Gap in Relationships?
Here’s a sweet dataviz! - Bavericks & Heretics - Ideas rejected, later proven correct
More data science links for you to read if you have a little time on your hands.
- Human-Level AI Is Right Around the Corner—or Hundreds of Years Away by IEEE Spectrum
- Secret Algorithms Threaten the Rule of Law by MIT tech review
- Three Reasons ‘Wonder Woman’ Has Already Made History by FiveThirtyEight
- So, bots you say… by Alex Ramirez
- Learning to Cooperate, Compete, and Communicate by open.ai
- How big data can help you pick better wine by Quid
- Scientists slash computations for deep learning by phys.org
- DAWN: Infrastructure for Usable Machine Learning by Stanford
- Tensorflow I Love You, But You’re Bringing Me Down by Nate Harada
- Foundations for deep learning by Pauli Space
- Neural networks Exercises (Part-1) by Rexercises
- Google Brain Residency by Tiny Clouds
- the levity of serverlessness by tecznotes
- A Query Had To Go by activity.club
- Google’s AI-Building AI Is a Step Toward Self-Improving AI by SingularityHub
- Why Apple is struggling to become an artificial-intelligence powerhouse by Washington Post
About a month ago Google open sourced the largest dataset of doodles from their Quick Draw! app, for people to play with. As planned, people are now starting to release some pretty cool stuff!
So first up is a project organising the doodles by similarities, so you can scroll through how people draw dogs.
Up next we have a project by Quartz analysing how different cultures draw circles (and where you fit in).
And a team favorite is the Forma Fluens project with 3 different visualisations of the data, including this video superimposing drawings by people around the world, to show how similarly people around the world think.
Face ML - Indexing faces on Instagram
Sciency Data Viz - Cracking the mystery of egg shape
Ai for dummies - Grandma should understand Artificial Intelligence. Here’s how I explained her
Cybersecurity - The RNC Files: Inside the largest US voter data leak
90’s Memorabilia - Here’s a random goodie: A map of all the area codes where Luda Cris has his hoes
———–
Want to get next week’s links? Sign-up to the newsletter now!
———–
As always, we didn’t have enough room to put all the stuff we’re reading in this week’s newsletter - so here’s the rest.
- A poor man’s video analyzer by Longhow Lam
- Gradient Boosting – the coolest kid on the machine learning block by DisplayR
- A Glimpse Under the Hood of Natural Language Processing by Lytics
- Deep Learning With The Beast by Tooploox
- 10 Algorithms Every Programmer Should Know – and When to Use Them by Professor Jake
- How to catch the Saimaa ringed seal? — Story about applying modern deep learning to computer vision by Emblica
- Python Plotting for Exploratory Analysis by tdhopper
- Google Datalab for Machine Learning Education by Google
- How to start with Machine Learning by MLjar
- An introduction to Support Vector Machines (SVM) by MonkeyLearn
- [Video] A famous venture capitalist predicts big banks will fall first to artificial intelligence
- [Awesome] visualising data structures and algorithms through animation
- New algorithm generates practical paper-folding patterns to produce any 3-D structure.
- [Help Out] Help EFF Track the Progress of AI and Machine Learning
- [Research] What works in e-commerce - a meta-analysis of 6700
online experiments
- What Kind of Dog Is It - Using TensorFlow on Mobile Device by Jeff Tang
- Should Uber’s next CEO be a robot? by Rough Type
“We’re in an awkward phase of the tech industry, one marked by incremental improvements to technologies that we think of as boring — and lots of exciting promises about far-off tech that isn’t quite ready for prime time.” NY Times
Google made AI headlines this week with all the new features announced at this year’s Google I/O. Unsurprisingly, many of the (relatively small) innovations presented rely on its advances in deep learning:
- With Google Lenses you can take a picture and your phone will tell you what is in it
- It will also remove unwanted objects from your pics;
- Gmail now comes with smart responses
- And Google Assistant wants to take over iPhones
- And … drumroll… Smart Copy-Pasting!!
The most important presentation though was the TensorFlow Research Cloud, which leverages Google’s TPU chip to create a platform for machine learning research.
Our favorites this week
Awesome Dataviz: Are Pop Lyrics Getting More Repetitive?
DIY: NotHotdog-Classifier
Becoming Data Driven: Democratizing Data at Airbnb
Research of the Week: We Recorded VCs’ Conversations and Analyzed How Differently They Talk About Female Entrepreneurs
Here’s an AI goodie! A video of Robots learning to do tasks
Here are some other things we’re reading this week.
- GDPR: Challenges and Strategies for Compliance - by Dataiku
- Whose Parents Pay for College? by Pricenomics
- Logs and Metrics - by Cindy Sridharan
- Eliminating the Human - by David Byrne
- Questions & Intuition for Tackling Deep Learning Problems - by the Ecommercer Intelligencer
- Hard-earned advice for AI products - by Josh Winters
- We Don’t Need No Stinkin’ Databases - by Bill Torpey
- Big Data for Big Business? A Taxonomy of Data-Driven Business Models used by Startups - by Antoine Buteau
- [Resource] Releasing the World’s Largest Street-level Imagery Dataset for Teaching Machines to See - by Mapillary
- [Paper] Checking How Fact-Checkers Check - by Chloe Lim
- [PDF Book] Street-Fighting Mathematics - by Sanjoy Mahajan
- Building a smarter Hacker News - by Techcrunch
Fresh research has revealed that those organizations leading the pack when it comes to data science (those using many different data sources and developing highly qualified data science teams) are 4x more likely to create revenue growth that surpasses shareholder expectations.
That’s not all - many smaller companies appear to be driving data science innovation.
Read the full article by Bernard Marr
- Big Data Survey
- What I’ve learned about neural network quantization by Pete Warden
- Predictive learning vs. representation learning by Harvard Intelligent Probabilistic Systems
- [Presentation] Automating Data analysis Pipelines
- Zero to One — A Ton of Awe-Inspiring Deep Learning Demos with Code for Beginners by Sam Putnam
- Watch 35 Years of the World’s Economy Evolving as a Living Organism by howmuch
- [Data viz] Alone Time by Nathan Yau
- [Paper] Two Decades of Recommender Systems at Amazon.com
- A generative model for music track playlists by IHeartRadio
- What tomorrow’s business leaders need to know about Machine Learning?by Bill Schmarzo
- [Video] China’s All-Seeing Surveillance State Is Reading Its Citizens’ Faces by the Wall Street Journal
Never enough room in this week’s newsletter to share all the links we love. Here are a few extra!
“Meet the classified artificial brain being developed by US intelligence programs”
A long exposé about the fairly secret project Sentient - an artificial brain developed by the National Reconnaissance Office.
Check out the article by the Verge