— In the News —
Justices of the U.S. Supreme Court recently dismissed a quantitatively reasoned argument as "gobbledygook." That particular dismissal isn't used very often but documentation of the court’s math problem fills pages in academic journals. This article by FiveThirtyEight explores how some of the most powerful jurists in the world are afraid of math.
Remember AlphaGo? It made huge news last year when it defeated a human grandmaster. Now, AlphaGo Zero takes that up yet another notch by teaching itself how to play Go from scratch. Not only that, in a head-to-head match with the original AlphaGo, this latest version defeated the original, 100 games to zero.
Just as the World Wide Web connects documents, the web of the future will connect data. It will be a web of linked data and for some, this Semantic Web is already here.
— Sponsored Link —
Mixpanel recently aggregated and analyzed behavior data for 1.3 billion unique users who triggered 50 billion events in SaaS, media, finance, or e-commerce & retail products. They found median and 90th percentile benchmark metrics for core KPIs like growth rates, retention and conversion. Get the free report now.
— Tools and Techniques —
Engineers commonly develop software using "Agile" methodologies, which, among other things, enables them to develop robust software, efficiently. Here's a good overview of Agile and how it directly applies to data science.
Sebastian Ruder's latest post looks at the deficiencies of pre-trained word embeddings and considers recent approaches for trying to resolve them. Includes lots of linked references for further exploration.
Data that arrives in a continuous, never-ending stream needs to be handled differently than data that's known upfront. This post describes an experimental library that enables you to handle things like streaming time series data and IoT telemetry using Pandas and Streamz. This looks super useful and is being actively developed.
In his latest post, Peter Ellis explores half a million rows of disaggregated crash data for New Zealand, and along the way illustrates geo-spatial projections, maps, forecasting with ensembles of methods, a state space model for change over time, and a generalized linear model for understanding interactions in a three-way cross tab.
— Deep Learning —
Tryolabs recently launched an open source toolkit for computer vision. It currently supports object detection and image classification and there's a lot more to come. This post describes the project, the problems it's solving, and where it's going.
iPhone users can say, "Hey Siri" and the phone will magically wake up and wait for a verbal command. It's a simple interface but there's a lot going on behind the scenes. In this 6th issue of Apple's Machine Learning Journal, the Siri Team explains how it works.
— Conferences —
The Data Natives Conference is billed as "the meeting point for industry experts, entrepreneurs, tech and business professionals to inspire each other and disrupt the status quo." It looks like a great conference and if you're interested in going, Data Elixir readers save 30% with this code: DNLOVESDA30
— Career —
There are a lot of worthwhile ideas here that go well beyond the typical advice about which skills to learn. Here's how to identify and develop your personal critical path, including considerations for selecting workplaces, projects, and mentors.
— In Case You Missed It —
Be sure to catch the most popular links from last week's issue...
— About —
Data Elixir is curated and maintained by @lonriesberg. If some awesome person forwarded this issue to you, subscribe for free at dataelixir.com and get it delivered every week.