Data Science newsletter – July 5, 2017

Newsletter features journalism, research papers, events, tools/software, and jobs for July 5, 2017

GROUP CURATION: N/A

 
 
Data Science News



Who Will Win the IoT Platform Wars?

Boston Consulting Group; Akash Bhatia, Zia Yusuf, David Ritter and Nicolas Hunke


from

More than 400 companies offer IoT platforms today. Enterprise software and service companies and IoT startups account for the largest portion (22% and 32%, respectively) of companies that claim to offer IoT platforms. In addition, industrial technology providers (at 18%) are offering IoT platforms in an effort to shift away from a hardware-centric business model. Internet companies and telcos make up the remainder of the IoT platform vendors.

Although some companies are emerging from the pack as possible leaders, choosing the right provider from the plethora of companies vying for the platform market remains a challenge. Buyers are looking for insight into how to make the best choice when selecting an IoT platform today. The most useful approach, we believe, includes analyzing the ecosystem from the vendor’s perspective and identifying the key factors that will determine which companies win the IoT platform wars.


By sharing camera trap data we can monitor wildlife status globally

Remote Sensing in Ecology and Conservation blog, Timothy G. O’Brien


from

Article 29 of the Nagoya Protocol mandates all signatories to the Convention on Biological Diversity (CBD) to monitor their implementation of CBD obligations and document progress toward Aichi 2020 targets. Given the multi-dimensional character of biodiversity, a single, comprehensive metric is clearly not feasible. Rather we rely on a range of factors to measure the status and conservation of biodiversity. Pereira et al, (2013) proposed a list of essential biodiversity variables (EBVs) to enable the study, reporting and management of biodiversity change.

Three EBVs of particular interest to wildlife biologists and conservationists are species distribution, population abundance, and taxonomic diversity. Global trend dashboards using such data include the Wild Bird Index, the Living Planet Index and the Wildlife Picture Index. The Wildlife Picture Index uses data from camera trap images to generate trends in species abundance and distribution. Camera trap data can, of course, also be used to assess trends in species composition and taxonomic diversity. Automated camera traps remotely sense the passage of moderate sized wildlife, are especially useful for monitoring terrestrial and semi-terrestrial mammals and birds, and are, therefore, an excellent method for gathering data on EBVs.


Humans, Machines and the Future of Predictive Maintenance

IIOT World


from

There’s no denying that advancements in artificial intelligence and machine learning are occurring faster than most experts anticipated. The era of self driving cars is just on the horizon, the Jabberwacky chatbot can convince you it’s a human, and Google’s deep dream can produce creative works of art. Advancements in networking, information processing algorithms, and data storage technologies are enabling computers to acquire complex skillsets and capabilities. In the meantime, the world is left to wonder exactly how these technologies will be implemented and how they will impact existing markets and industries. The predictive maintenance industry is no exception. There is no question that predictive maintenance is a superior strategy in comparison to common preventive maintenance and especially reactive maintenance. According to the Department of Energy’s operations and maintenance best practices guide, a predictive maintenance strategy can realize savings of 30-40% and 8-12% over reactive and preventive strategies respectively.

The benefits of a predictive program originate from the fact that it is a strategy driven by data such as vibration, ultrasound, temperature, electrical, and other measurements. This data ultimately drives informed decisions such as figuring out the specific repair required to fix a machine or timing the repair appropriately to minimize the risk of catastrophic failure.


Why Artificial Intelligence is Different from Previous Technology Waves

Unsupervised Methods, Robbie Allen


from

I’ve been around computing since my older brother got a Commodore 64 for Christmas in 1983. I took my first “business machines” class in high school in 1991, attended my first computer science class in 1994 (learning Pascal), and moved to Silicon Valley in 1997 after Cisco converted my internship into a permanent position. I worked in Cisco’s IT department for several years before moving to their engineering group where I designed networking protocols. I went to grad school at MIT in 2004 where I met the founders of several companies in Y Combinator’s first couple of batches and worked on Hubspot before it was Hubspot. After writing several books for O’Reilly and attending the first O’Reilly Web 2.0 and MIT Sloan Sports Analytics conferences, I started a “Web 2.0 for Sports” company called StatSheet.com in 2007, which in 2010 pivoted into the first Natural Language Generation (NLG) company called Automated Insights. I recently stepped back at Ai to become a Ph.D. student at UNC studying Artificial Intelligence.

All of that to say I’ve had a bird’s eye view to watch the incredible innovation that’s occurred over the past 30 years in technology. I’ve been lucky to be in the right place at the right time.


Local startups need to make noise about voice computing

The Boston Globe, Scott Kirsner


from

At a demo showcase held at District Hall in South Boston last week , everyone seemed excited about the potential of voice-driven technology. A local Amazon employee, Robert McCauley, talked about how easy it is to build games and other apps for the Echo device. He also plugged a new device, the Echo Show, which includes a screen for conducting video chats with friends or seeing who’s standing at your front door.

The showcase also included entrepreneurs like Scott Cohen of BigR.io, a Concord consulting firm that is creating its own intelligent persona, named Jaxon, to help retailers conduct conversations with prospective customers, answer questions, and — ideally — close the deal.

Boston-based Vesper Technologies was showing low-power microphones that could be built into all sorts of devices — from smartphones to speakers to trash cans — founder Matt Crowley explained, so that you can control them with your voice.


Vision Zero Labs: Using Data Science to Improve Traffic Safety

Microsoft New York


from

The central idea behind the global Vision Zero movement is that traffic crashes are preventable.

At Microsoft, we believe that data science and complex machine learning can aid cities in their life-saving Vision Zero commitment. That’s why we partnered with Datakind in 2015, and since then, we’ve worked with them to use city-specific data to identify where traffic safety conditions could be improved to ease traffic and protect citizens.

Today, we are releasing this video case study to showcase the project, its learnings, and its future potential.


Generation CS Drives Growth in Enrollments

Communications of the ACM, blog@ACM, Mark Guzdial


from

The new Computing Research Association (CRA) report “Generation CS: Computer Science Undergraduate Enrollments Surge Since 2006” (http://cra.org/data/Generation-CS/) describes the dramatic increase in enrollments in computer science (CS) over the last 11 years, with an especially rapid increase since 2009. Sixty percent of academic units surveyed more than doubled their enrollment in that time. The report describes a new generation of undergraduate students who realize the importance of computing education.


Machine Learning Comes to Tour de France

insideBIGDATA


from

Amaury Sport Organisation (A.S.O.), organizers of the Tour de France, and Dimension Data, the Official Technology Partner of the Tour de France, announced the introduction of machine learning technologies at this year’s Tour de France to give cycling fans across the globe an unprecedented experience of this year’s event. The race begins in Düsseldorf on Saturday and finishes at the Champs-Elysees in Paris on 23 July.

This year, Dimension Data’s data analytics platform, which was developed in partnership with A.S.O., incorporates machine learning and complex algorithms that combine live and historical race data to provide even deeper levels of insight as the race unfolds. Fans will also benefit from rider profiles to understand more about environments and circumstances in which riders perform best.

As part of a new pilot this year, A.S.O. and Dimension Data are exploring the role of predictive analytics technologies to assess the likelihood of various race scenarios, such as whether the peloton will catch the breakaway riders at certain stages of the race.


By teaching computers to track asteroids, UW scientists may save the Earth

The Seattle Times, Katherine Long


from

In five years, a sky-scanning telescope in Chile will begin hunting the heavens for asteroids on a collision course with Earth, and scientists at the University of Washington are at the forefront of work to spot them.


Kernel-predicting Convolutional Networks for Denoising Monte Carlo Renderings

Disney Research


from

We introduce a deep learning approach for denoising Monte Carlo-rendered images that produces high-quality results suitable for production. We train a convolutional neural network to learn the complex relationship between noisy and reference data across a large set of frames with varying distributed effects from the film Finding Dory. The trained network can then be applied to denoise new images from other films with significantly different style and content, such as Cars 3 (right), with production-quality results.


Company Data Science News

Jawbone is shutting down, but it’s founder Hosain Rahman, is pivoting to Jawbone Health, taking many of Jawbone’s employees with him. Jawbone’s software will be maintained by the new company and you can bet that this pivot is all about the wealth available in the health data the devices have been collecting.

SoundCloud laid of 173 people and is closing its London and San Francisco offices while employees in the New York and Berlin offices fight off bankruptcy.

Disney is using deep learning to denoise frames from animated movies. They used Finding Dory as a training set and then applied the model to Cars 3 with robust results. Yes, it’s possible I included this story because I wanted to be able to write about Finding Dory and Cars 3. But seriously, Disney is a bigger player in image recognition than most realize; I have a responsibility to reveal this reality.

Peter Norvig‘s latest proclamation can be taken straight to your boss (if you work in AI). Gently explain to him/her/they that micro-management is out, teaching is in. As an organizational sociologist, I am always here to support your efforts to oppose micromanagement, but 1) I do not have the clout Peter Norvig does and 2) I have never provided evidence that the robots don’t like micro-management. I usually think in terms of human workers, but maybe that is passé? If you try to resist micro-management to ensure the proper care and training of your AI, please report back.

Google bought Kaggle and now the platform is running a $1.5m competition that’s only open to American competitors due to the nature of the data (US airport screening data). Is this a sign of an organizational culture clash between the more radically meritocratic Kaggle and Google’s play-within-the-lines corporate gigantism?

NVidia announced another powerhouse partnership, this time with Baidu. Yep, it’s still primarily about self-driving cars using NVidia’s kick-a$$ GPU chips.

Now that Facebook has 2 billion users – which is a kind of growth ceiling on the new user front – Zuckerberg and friends are going to try to get more users into Facebook groups. Of course, the groups will now be suggested by recommender systems. It’s a bit of a head scratcher why groups weren’t already being served up by helpful bots….shrug.

DeepMind could get access to 100,000 genome sequences in the UK, raising questions about how to proceed ethically and efficiently to bring the best health care to the largest number of people without introducing unintended consequences.

If you are thinking of launching a startup, don’t launch AI for parking. Your company will likely fail like these and the reason is a fairly straightforward combination of econ 101 and Newtonian physics (if you can make a car orbit a parking spot the way electrons orbit a nucleus, you may have something). This is your weekly lesson in “AI does not solve all the problems.”

But AI could help block terrorists from banking infrastructure which seems more important than getting a good parking spot.

I love a company history! Here’s a story about the sleep-deprived race to win the Netflix prize.



Amazon Prime is on pace to become more popular than cable TV. Innovate or die. And reality TV does not count as an innovation anymore. For real. 🙂

How does Instagram‘s recommender work? Honestly…not by using image recognition, but by relying on all that excellent metadata people are using. What? No wonder it doesn’t work, #peopledon’tknowhowtousehashtags.

Using clarifai as a partner, West Elm is using image recognition and pinterest to recommend furniture. I bet this works WAY better. Zuckerberg, call clarifai immediately. Or, you know, buy clarifai immediately. [NB: Zuckerberg does not subscribe to this newsletter.]


We need ethical safeguards to stop our brains getting hacked, say the experts

TechRadar, Andrew London


from

We are at a very early stage with these technologies and a group of the world’s leading neuroscientists, including Dr Birbaumer, are calling for ethical guidelines to be implemented now. According to Jens Clausen from the Center for Ethics in the Sciences at the University of Tubingen:

“Technological advances in the BMI field are currently developing at such rapid rate that it is high time to define a legal and ethical framework”


Duke Researchers Find a Shortcut to Predicting New Magnetic Materials

Design News, Tracey Schelmetic


from

Materials scientists from Duke University have demonstrated a shortcut to the traditional trial-and-error process. Using high throughput computational models that predict magnetism in new materials, the scientists have successfully developed, atom by atom, two new magnetic materials: cobalt, magnesium and titanium (Co2MnTi); and manganese, platinum and palladium (Mn2PtPd).

Using the computer model, the researchers focused on Heusler alloys, or materials made with atoms from three different elements arranged in one of three different structures. With 55 elements to choose from (and all possible potential arrangements), the manual process would have required testing 236,115 combinations. The model permitted the team to test hundreds of thousands of possibilities rapidly, resulting in two magnets that could be fabricated at thermodynamic equilibrium.


Big pharma turns to AI to speed drug discovery, GSK signs deal

Reuters, Ben Hirschler


from

The world’s leading drug companies are turning to artificial intelligence to improve the hit-and-miss business of finding new medicines, with GlaxoSmithKline unveiling a new $43 million deal in the field on Sunday.

 
Events



European Hackathon on Football

Italian Football Federation


from

Trento, Italy The event is organised by the Italian Football Federation, in collaboration with the University of Trento and the Autonomous Province of Trento. October 14-15. [registration required]

 
Tools & Resources



Official Python interface to CoreNLP using a bidirectional server-client interface.

GitHub – stanfordnlp/python-stanford-corenlp


from

This package contains a python interface for Stanford CoreNLP that contains a reference implementation to interface with the Stanford CoreNLP server. The package also contains a base class to expose a python-based annotation provider (e.g. your favorite neural NER system) to the CoreNLP pipeline via a lightweight service.


Deep Learning on ROCm

ROCm


from

Announcing our new Foundation for Deep Learning acceleration MIOpen 1.0 which introduces support for Convolution Neural Network acceleration — built to run on top of the ROCm software stack!


Visualizing Spanish Migration: A Case Study for Data Exploration at Periscopic

Periscopic, Wes Bernegger


from

Recently, one of our founders, Dino Citraro, attended the Rostock Retreat in Germany as a keynote speaker. Prospective participants were asked to visualize one of three demographic datasets as part of their application. In preparation for his talk we decided to play around with a couple of those datasets as case studies of how we explore and visualize data.

 
Careers


Postdocs

Postdoctoral Fellow – Urban Genome Project



University of Toronto; Toronto, Canada
Full-time positions outside academia

Technical Project Lead for Open Algorithms (OPAL) Project



Overseas Development Institute; London, England

Leave a Comment

Your email address will not be published.