Web Picks (week of 11 December 2017)

Every two weeks, we find the most interesting data science links from around the web and collect them in Data Science Briefings, the DataMiningApps newsletter. Subscribe now for free if you want to be the first to get up to speed on interesting resources.

  • Using Artificial Intelligence to Augment Human Intelligence
    By creating user interfaces which let us work with the representations inside machine learning models, we can give people new tools for reasoning.
  • What Happens When the Government Uses Facebook as a Weapon?
    It’s social media in the age of “patriotic trolling” in the Philippines, where the government is waging a campaign to destroy a critic—with a little help from Facebook itself.
  • Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm (paper)
    DeepMind put itself at the forefront of the news again this week by beating the leading chess engine, StockFish, using the same architecture as was used by AlphaZero to beat the Go world champion.
  • Turi Create simplifies the development of custom machine learning models.
    This was a surprise — after being acquired by Apple, we didn’t think that we’d see the source code behind Graphlab Create’s machine learning framework (later: “Dato”, later: “Turi”) being available again, but here we go: “Turi Create simplifies the development of custom machine learning models. You don’t have to be a machine learning expert to add recommendations, object detection, image classification, image similarity or activity classification to your app.”
  • Apache Kafka and GDPR Compliance
    “Our commitment is to provide the necessary capabilities in data streaming systems, to allow your data-driven business to achieve compliance with GDPR prior to the regulation’s effective date.”
  • A Year in Computer Vision
    Very informative wrap-up article on computer vision and recent improvements in the field.
  • Artwork Personalization at Netflix
    “Artwork may highlight an actor that you recognize, capture an exciting moment like a car chase, or contain a dramatic scene that conveys the essence of a movie or TV show. If we present that perfect image on your homepage (and as they say: an image is worth a thousand words), then maybe, just maybe, you will give it a try.”
  • Want a loan? Make sure you’re tweeting the right things
    The article that someone tweeted about, posts that they liked on Facebook, and a new phone just bought on an e-commerce site—all these events now play a crucial role in determining if an individual is eligible for a loan or not.
  • Google collects Android users’ locations even when location services are disabled
    Many people realize that smartphones track their locations. But what if you actively turn off location services, haven’t used any apps, and haven’t even inserted a carrier SIM card?
  • Analyzing 1000+ Greek Wines With Python
    Another interesting approach of web scraping!
  • A Massive New Library of 3-D Images Could Help Your Robot Butler Get Around Your House
    Using three-dimensional images is a better way of mimicking the way animals perceive things.
  • Optimization for Deep Learning Highlights in 2017
    Deep Learning ultimately is about finding a minimum that generalizes well — with bonus points for finding one fast and reliably. Our workhorse, stochastic gradient descent (SGD), is a 60-year old algorithm, that is as essential to the current generation of Deep Learning algorithms as back-propagation.
  • Neural Networks in JavaScript with deeplearn.js
    Yes, GPU support is included through pixel/vertex shaders on WebGL, impressive stuff!
  • Monte Carlo Simulation with Categorical Values
    In Monte Carlo simulation, we repeatedly make guesses of some unknown value according to some distribution and are able to report on the results of that simulation to understand a little bit more about the unknown. While any one guess may be far from the truth, in aggregate those outliers don’t have as much of an effect.
  • Are GANs Created Equal? A Large-Scale Study (paper)
    “We conduct a neutral, multi-faceted large-scale empirical study on state-of-the art models and evaluation measures. We find that most models can reach similar scores with enough hyperparameter optimization and random restarts. This suggests that improvements can arise from a higher computational budget and tuning more than fundamental algorithmic changes.Finally, we did not find evidence that any of the tested algorithms consistently outperforms the original one.”
  • Neural Network on a Commodore 64
    Who says neural networks are new? Take a trip back to 1987 and find out!