Data Science newsletter – April 22, 2019

GROUP CURATION: N/A

On the east side of Bryant Park in Midtown Manhattan, three cameras on
the roof of a restaurant film the lunch crowds, tourists and commuters —
everything that goes on each day. The feeds are streamed publicly online.

American Family Insurance said Friday it will invest $20 million in data science initiatives at the University of Wisconsin-Madison, expanding an existing partnership between the insurer and university.

The award includes $10 million for research over the next decade and a $10 million endowment to create the American Family Insurance Data Science Institute on campus.

WHO today released new recommendations on 10 ways that countries can use digital health technology, accessible via mobile phones, tablets and computers, to improve people’s health and essential services.

“Harnessing the power of digital technologies is essential for achieving universal health coverage,” says WHO Director-General Dr Tedros Adhanom Ghebreyesus. “Ultimately, digital technologies are not ends in themselves; they are vital tools to promote health, keep the world safe, and serve the vulnerable.”

Over the past two years, WHO systematically reviewed evidence on digital technologies and consulted with experts from around the world to produce recommendations on some key ways such tools may be used for maximum impact on health systems and people’s health.

MD Anderson Cancer Center is ousting three scientists in connection with concerns China is trying to steal U.S. scientific research, the first such publicly disclosed punishments since federal officials directed some institutions to investigate specific professors in violation of granting agency policies.

MD Anderson took the actions after receiving e-mails last year from the National Institutes of Health, the nation’s largest public funder of biomedical research, describing conflicts of interest or unreported foreign income by five faculty members. The agency, which has been assisted by the FBI, gave the cancer center 30 days to respond.

Researchers have begun sampling water and sediment to understand the extent of the problem

More than double the number of trips were taken in 2018 than the year prior on shared micromobility, a fast-growing and rapidly-evolving form of transportation in the United States

The National Association of City Transportation Officials (NACTO), an association of 69 major North American cities, today released its annual comprehensive count of all shared micromobility (shared bike and e-scooter) trips in the United States.

Before computers, no sane person would have set out to count gender pronouns in 4,000 novels, but the results can be revealing, as MIT’s new digital humanities program recently discovered.

Launched with a $1.3 million grant from the Andrew W. Mellon Foundation, the Program in Digital Humanities brings computation together with humanities research, with the goal of building a community “fluent in both languages,” says Michael Scott Cuthbert, associate professor of music, Music21 inventor, and director of digital humanities at MIT.

“In the past, it has been somewhat rare, and extremely rare beyond MIT, for humanists to be fully equipped to frame questions in ways that are easy to put in computer science terms, and equally rare for computer scientists to be deeply educated in humanities research. There has been a communications gap,” Cuthbert says. “That’s the genesis of this new approach to computation in humanities.”

Facebook is working on a voice assistant to rival the likes of Amazon’s Alexa, Apple’s Siri and the Google Assistant, people familiar with the matter told CNBC.

The effort is coming out of the company’s division that works on long-term tech projects and hardware, including the company’s virtual reality Oculus headsets.

A team based out of Redmond, Washington, has been spearheading the effort to build the new AI assistant.

When Jillian York, a 36-year-old American activist, was on vacation in February, she received an unexpected text. Her friend Adam Harvey, another activist and researcher, had discovered photos of her in a US government database used to train facial-recognition algorithms, and wondered whether she knew about it.

York, who works in Berlin for the Electronic Frontier Foundation, a digital rights non-profit group, did not. She was stunned to discover that the database contained nearly a dozen images of her, a mixture of photos and YouTube video stills, taken over a period of almost a decade.

When she dug into what the database was used for, it dawned on her that her face had helped to build systems used by the federal government to recognise faces of interest, including suspected criminals, terrorists and illegal aliens.

“What struck me immediately was the range of times they cover,” York says. “The first images were from 2008, all the way through to 2015.” Two of the photos, by a photographer friend, had been scraped from Google. “They were taken at closed meetings. They were definitely private in the sense that it was me goofing around with friends, rather than me on stage,” she adds.

Deep learning takes a completely fresh approach to determining consumer response, in three key ways:

1 . First, it does not rely on a single, easy-to-interpret equation, but rather on a series of linear and non-linear transformations each representing a “neural” layer that is linked to the next layer. When there are many intermediate (hidden) layers, the method is called deep learning. DL models often require large data bases to estimate the many parameters.

2. Secondly, DL methods are judged by their ability to make accurate predictions from a new data base not just to fit an existing model. A DL model is good if it predicts well.

3. The third difference is structural advantage. DL can handle high-dimensional data that classical methods cannot.

A team of researchers from the Dana-Farber Cancer Institute, Brigham and Women’s Hospital, Harvard Catalyst, Harvard Business School and the Laboratory for Innovation Science at Harvard collaborated with network and crowdsourcing platform Topcoder to run the 10-week prize-based contest calling on data scientists to develop AI-based solutions to address the task of tumor segmentation and replicate the accuracy of an expert radiation oncologist.

This paper applies information theoretic principles to the investigation of physical human-robot interaction. Drawing from the study of human perception and neural encoding, information theoretic approaches offer a perspective that enables quantitatively interpreting the body as an information channel and bodily motion as an information-carrying signal. We show that ergodicity, which can be interpreted as the degree to which a trajectory encodes information about a task, correctly predicts changes due to reduction of a person’s existing deficit or the addition of algorithmic assistance. The measure also captures changes from training with robotic assistance. Other common measures for assessment failed to capture at least one of these effects. This information-based interpretation of motion can be applied broadly, in the evaluation and design of human-machine interactions, in learning by demonstration paradigms, or in human motion analysis.

Launched by AICEP, the state agency for promotion of exports and investment, the platform uses AI technology, including machine learning, big data and design thinking to deliver customized services to thousands of companies.

“The platform will be a brave new world,” AICEP’s president Luis Castro Henriques told Reuters. “It will allow us to attract more companies to internationalize, serve companies better and be more productive.”

Called “Portugal Exporta”, it offers a range of services to clients, such as matching between companies and investors, information on potential partners and customized internationalization plans for each firm.

Amid a torrid geopolitical, commercial and scientific race around artificial intelligence, universities are adding professors, classes and entire new programs, but there is still a massive talent shortage, forcing companies to contemplate creative ways around it.

The big picture: The frenzy at American and Canadian universities reflects the changing technology cycle, in which AI is expected to become perhaps the defining factor in economic and geopolitical power in the decades ahead.

Students are pouring into computer science programs from coast to coast in the U.S. and Canada, university professors tell us. But the AI students among them still number at most in the low thousands in all at the moment, while companies say they are prepared to hire tens, if not hundreds, of thousands of AI experts.

the DoD moves forward in integrating AI and machine learning across its functions.
https://innovation.defense.gov/ai/

Pittsburgh, PA July 23-26. “ACH is the United States-based constituent organization in the Alliance for Digital Humanities Organizations (ADHO). The ACH2019 conference, in partnership with Keystone DH, provides a forum for conversations on an expansive definition of digital humanities in a broad array of subject areas, methods, and communities of practice.” [$$$]

Here we describe MorphNet, a sophisticated technique for neural network model refinement, which takes the latter approach. Originally presented in our paper, “MorphNet: Fast & Simple Resource-Constrained Structure Learning of Deep Networks”, MorphNet takes an existing neural network as input and produces a new neural network that is smaller, faster, and yields better performance tailored to a new problem. We’ve applied the technique to Google-scale problems to design production-serving networks that are both smaller and more accurate, and now we have open sourced the TensorFlow implementation of MorphNet to the community so that you can use it to make your models more efficient.

Generative models can seem like a magic box where you plug in observed data, turn some dials, and see what the computer spits out. SpaceSheet is a simple spreadsheet interface to explore and experiment for a clearer view of the spaces between. Even if you’re not into this research area, it’s fun to click and drag things around to see what happens.

Currently there is no standard way to identify how a dataset was created, and what characteristics, motivations, and potential skews it represents. To begin to address this issue, we propose the concept of a datasheet for datasets, a short document to accompany public datasets, commercial APIs, and pretrained models. The goal of this proposal is to enable better communication between dataset creators and users, and help the AI community move toward greater transparency and accountability. By analogy, in computer hardware, it has become industry standard to accompany everything from the simplest components (e.g., resistors), to the most complex microprocessor chips, with datasheets detailing standard operating characteristics, test results, recommended usage, and other information. We outline some of the questions a datasheet for datasets should answer. These questions focus on when, where, and how the training data was gathered, its recommended use cases, and, in the case of human-centric datasets, information regarding the subjects’ demographics and consent as applicable. We develop prototypes of datasheets for two well-known datasets: Labeled Faces in The Wild and the Pang \& Lee Polarity Dataset.

NVIDIA’s Isaac SDK and Isaac Simulator, which were announced last month, are now available for robotics developers to download.

The Isaac SDK toolbox offers developers access to Isaac applications, GEMs (robot capabilities), a Robot Engine, and the Isaac Sim. NVIDIA’s goal with its Isaac portfolio is to make it easier for manufacturers, researchers, and startups to add AI for perception, navigation, and manipulation into next-generation robots.