Belgium's New Data Companies, 2015's Tech Trends and Privacy-friendly Beacons (#WIC)

By December 15, 2014 Week in Context No Comments

This week in context, we are proud to say we were featured as one of Belgium’s New Data Companies, and have a look at 2015’s upcoming tech trends, learn about beacons and privacy and feature an illustrated guide to computer vision.

This Week in Context

Your Weekly Update on All Things Context, December 15 2014

Argus Labs was featured as one of Belgium’s ‘New Data Companies’ in De Tijd’s BNK (Perspectives on Banking) magazine. We’re described as having the most abstract but also most interesting product. Other data companies to keep an eye on are Octopin, Wigoh and Datacamp. Read the full article in Dutch here.


“As it stands, a pigeon brain can still outperform even the world’s most advanced neural nets, the Google Brain included, on some tasks. But Hinton joined Google, he says, to build one of the biggest neural networks to date, one that can “do pretty good common sense.”

Enabling Humanitarian Use of Mobile Phone Data
Yves-Alexandre de Montjoye, Jake Kendall and Cameron F. Kerry

Five Must-Reads

1. Trends to watch in 2015: Algorithmic Accountability, Deep Learning, Smart Cameras, .. 

The Wall Street Journals highlights Trends to Watch in 2015, less-noticed strategy shifts among tech companies, developments moving out of the lab and into the commercial world and some ‘intriguing metathemes of the year ahead’. The list reads like your average Week in Context: algorithmic accountability, smart cameras that can understand emotion, personal assistant apps and privacy, deep learning and ambient proximity using beacons.

Check out the WSJ’s pick from Webbmedia’s 2015 Trends Report on Digits. The full list of 2015 Tech Trends is also available on Slideshare.

2. Beacons are privacy-friendly

“Beacons are sometimes misunderstood as instruments of “surveillance,” capable of tracking smartphone owners’ movements without their awareness. This is incorrect. Beacons transmit a low-power signal that can be picked up by nearby Bluetooth-enabled mobile devices, including smartphones. Beacons themselves don’t collect data. They broadcast short-range signals that can be detected by apps on mobile devices in close proximity to a beacon.” 

More on Beacon technology in the Future Of Privacy’s Understanding Beacons, A Guide To Beacon Technology. Hat tip to @BavoCranium.

3. Computers learn to ‘glimpse’ at images

People recognise objects by selectively focusing on the important parts of an image instead of processing an entire image at once. Computers are being thought to do this too. By ignoring irrelevant noisy features in an image, fewer pixels need to be processed, substantially reducing classification and detection complexity.

Google DeepMind presented Recurrent Models of Visual Attention, a paper which describes an “attention-based task-driven visual processing” that is capable of extracting information from an image or video by adaptively selecting a sequence of smaller regions (glimpses), processing only selected regions at high resolution. 

Read on at Research at Google’s G+ stream. Hat tip to our data scientist Joren, who highlights a neat example where the algorithm is used to play Pong.

4. What smartphones can mean for weather forecasting

Unlike traffic data, however, there are significant restraints on the type of data that can be crowdsourced for weather prediction. Temperature data, for example, is unreliable; ambient temperatures may be 90°Fahrenheit or -20° Celsius, but if a phone feeding data is in an air-conditioned car or a heated house, the data it yields is useless. According to Mass, the most reliable data to seek from a mobile-enabled network is barometric pressure; “every other parameter is useless, it makes no sense,” he says. “Pressure is absolutely unique.”

That’s Traffic; Up Next, Weather on reports that PressureNet is currently collecting data from about 40,000 nodes, and is looking to OpenSignal but also AccuWeather and The Weather Network to expand upon their ‘sensing base’.

5. Netflix’ Atlas: time-series telemetry

Netflix has released details and an API for Atlas, the centralized monitoring system underlying the streaming service’s platform and ecosystem. Atlas allows Netflix engineers to send and query time-series data, scaled over time to handle the rapidly growing Netflix user base. They are open-sourcing the query layer and plan to open source more of the ecosystem ‘as soon as feasibly possible’.

A work of Art vs Tech: An Illustrated Guide To Computer Vision

Screen Shot 2014-12-15 at 21.55.46It is a bit of a simplification, especially compared to the research mentioned above (glimpses) and the papers below.

However, Intel iQ’s “Through The Eyes of The Machine, an illustrated guide to computer vision” does explain ‘how computers see’ in 6 images; From Color by Numbers to Make educated guesses.

The guide on Medium.
Intel RealSense – iQ Special Edition


Papers, Talks & Research

  • Predicting next location using a variable order Markov Model (machine learning, location prediction, paper)
  • Discovering frequent ADL patterns from wearable accelerometers (activities of daily living, pattern discovery, machine learning, paper)
  • Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images (computer vision, deep neural networks, image recognition, paper)
  • SenseMe: A System for Continuous, On-Device, and Multi-dimensional Context and Activity Recognition (mobile computing, sensors, context-aware, paper)
  • Recurrent Models of Visual Attention (machine learning, pattern recognition, computer vision, paper)
  • No technical understanding required: Helping users make informed choices about access to their personal data (privacy, personal data, mobile, paper)
  • Bleep bleep!: determining smartphone locations by opportunistically recording notification sounds (context detection, audio, mobile computing, paper)
  • My Places Diary – Automatic Place and Transportation Mode Detection (pattern recognition, location, activity recognition, paper)
  • How transferable are features in deep neural networks? (machine learning, paper)

More interesting papers on mobile and ubiquitous computing are available on the European Union Digital Library, where the proceedings from Mobiquitous, i-LOCATE and CSSWearable were just published.


Have a great and productive week!
(and if you haven’t done so yet, kindly consider subscribing to the Week in Context here.)


Leave a Reply

Contact us