Data Science newsletter – March 5, 2020

GROUP CURATION: N/A

The trick is a timely combination of side-to-side and see-sawing motions.

Responding to the evolution of the workforce, the University of Toledo will debut two new undergraduate degrees in data science and analytics to address what university administrators say is rising demand for data-savvy professionals.

In fall 2020, UT will offer a bachelor of arts degree in data analytics in the College of Arts and Letters, and a bachelor of science degree in data science in the College of Natural Sciences and Mathematics.

The Silicon Integration Initiative (Si2) has launched a special interest group to focus on the growing needs and opportunities in artificial intelligence (AI) and machine learning (ML) for electronic design automation.

The Si2 group will be chaired by researchers from IBM and Samsung and aims to identify current solutions and technology gaps in AI and ML strategies for EDA digital design.

“AI and ML are changing semiconductor design and improving performance and time to market,” said Leigh Anne Clevenger, Si2 design automation data scientist. “Based on member company interest, we expect the SIG to propose prototype projects to accelerate the development of standards in areas such as machine learning training, and data handling and sharing.

This year we received 917 proposals from ~50 countries and over 330 universities, and had the opportunity to increase our investment in several research areas related to Health, Accessibility, AI for Social Good, and ML Fairness. All proposals went through an extensive review process involving 1100 expert reviewers across Google who assessed the proposals on merit, innovation, connection to Google’s products/services and alignment with our overall research philosophy.

As a result of these reviews, Google is funding 150 promising proposals across a wide range of research areas, from Machine Learning, Systems, Human Computer Interaction and many more, with 26% of the funding awarded to universities outside the United States. Additionally, 27% of our recipients this year identified as a historically underrepresented group within technology. This is just the beginning of a larger investment in underrepresented communities and we are looking forward to sharing our 2020 initiatives soon.

If advertising wants to keep using big data sets filled with consumer information, they’ll have to do a better job protecting them.

And differential privacy is one of the most promising weapons in the privacy-protecting arsenal.

This week on The Big Story, senior editor Allison Schiff dishes on differential privacy. It might sound like an arcane academic concept – which it is – but she’ll give you the elevator pitch in a way that anyone can understand. [audio, 27:41]

Using more than 30 years of earth satellite images, scientists at Tufts University and the non-profit conservation organization Defenders of Wildlife have discovered that habitat loss for imperiled species in the U.S. over this period was more than twice as great on non-protected private lands than on federally protected lands. As wildlife face a host of survival threats ranging from habitat destruction to global climate change, the study, published today in Frontiers in Ecology and the Environment, provides evidence that federal land protection and listing under the U.S. Endangered Species Act are effective tools for stemming losses in species habitat.

Momus Analytics’ predictive scoring system is using race to grade potential jurors on vague qualities like “leadership” and “personal responsibility.”

Sure, after decades of listening, there is still no message. But with more data to sift through, and new technologies with superior search capabilities, odds of hearing from E.T. are rapidly improving. If the probability in the decade 2011–2021 were x percent, it’s going to be 1,000 times x in the following decade, says Andrew Siemion, director of the Berkeley SETI Research Center. (SETI stands for Search for Extra-Terrestrial Intelligence.)

The reason for E.T. optimism stems largely from several new projects in the works, enhanced with advanced methods for discerning an actual message hidden in the static of cosmic cacophony.

Siemion, speaking in Seattle on February 15 at the annual meeting of the American Association for the Advancement of Science, reported a new release of data from Breakthrough Listen, a major enterprise for recording radio signals from space. Now available for others to analyze, the data dump contains 2 petabytes of information (to store that much, you’d need 2,000 of today’s typical PCs with their puny 1 terabyte hard drives).

“Drawing on the on-going collaborative project ‘Towards Democratic Auditing’ carried out by the Data Justice Lab, in this talk Lina Dencik will engage with the question of advancing civic participation in a context of rapid technological and social transformation. Lina will discuss experiments in new democratic practices to ensure legitimacy, transparency, accountability and intervention in relation to data-driven governance and outline emerging terrains for developing citizen agency in a datafied society.” [video, 39:20]

The safety and security of critical information – whether it is sensitive intellectual property (IP), financial information, personally identifiable information (PII), intelligence insight, or beyond – is of vital importance. Conventional data encryption methods or cryptographic solutions, such as Advanced Encryption Standards (AES), translate data into a secret “code” that can only be decoded by people with access to a decryption key. These methods protect data as it is transmitted across a network or at rest while in storage. Processing or computing on this data however requires that it is first decrypted, exposing it to numerous vulnerabilities and threats. Fully homomorphic encryption (FHE) offers a solution to this challenge. FHE enables computation on encrypted data, or ciphertext, rather than plaintext, or unencrypted data – essentially keeping data protected at all times. The benefits of FHE are significant, from enabling the use of untrusted networks to enhancing data privacy. Despite its potential, FHE requires enormous computation time to perform even simple operations, making it exceedingly impractical to implement with traditional processing hardware.

Researchers from Canada’s University of Waterloo gathered more than 700,000 anonymous online journal entries written by over 67,000 users of a mobile mood tracking app. They then developed an AI computer model that could identify keywords within the text.

In this year’s QS Subject Rankings for computer science, 104 of the 601 schools featured are situated in the United States, with five of these in the top 10 for computer science worldwide (the most of any country). Outside of the top 10, there are plenty of top computer science universities available, including Princeton University (11th), University of California, Los Angeles (UCLA, 15th), and University of Washington (18th). Three schools, Columbia University, Cornell University and New York University (NYU), are tied in 19th place worldwide.

Credit: Getty Images

New research from the University of Washington finds that a natural aptitude for learning languages is a stronger predictor of learning to program than basic math knowledge, or numeracy. That’s because writing code also involves learning a second language, an ability to learn that language’s vocabulary and grammar, and how they work together to communicate ideas and intentions. Other cognitive functions tied to both areas, such as problem solving and the use of working memory, also play key roles.

“Many barriers to programming, from prerequisite courses to stereotypes of what a good programmer looks like, are centered around the idea that programming relies heavily on math abilities, and that idea is not born out in our data,” says Chantel Prat.

San Francisco is poised to pass a controversial proposition that would almost certainly limit further office space development in the city, perhaps pushing more tech companies and startups to set up their HQs elsewhere.

Prop E‘s passing, which seemed likely Wednesday afternoon following Tuesday’s election, ties office development approval to the city’s ability to meet affordable housing goals, something that the city and its developers haven’t proven themselves all that capable of doing in recent years. Amid skyrocketing rents and a homeless crisis, there have been ample concerns that the structures in the city are being overstressed, low and moderate income residents are being pushed out and that the influx of tech startups is exacerbating the problem.

We know, irrefutably, one thing about the coronavirus in the United States: The number of cases reported in every chart and table is far too low.

The data are untrustworthy because the processes we used to get them were flawed. The Centers for Disease Control and Prevention’s testing procedures missed the bulk of the cases. They focused exclusively on travelers, rather than testing more broadly, because that seemed like the best way to catch cases entering the country.

“An open source project shepherded by the Linux Foundation aims to accelerate data science curricula while benefitting from the contributions of students and teachers. OpenDS4All is funded by IBM and is being developed by the University of Pennsylvania. The effort would give educators free access to information needed to develop data science coursework. In return, successful approaches would be folded back into what project promoters call ‘constantly evolving and improving’ curricula.”

“A starter ‘curriculum kit’ includes a set of open source building blocks that could be used to launch data science programs. Based on the Python programming language, the tools and frameworks include code, documentation and data sets, organizers said.”

“A new study introduces a reproducible graph neural network (GNN) benchmarking framework to study and quantify the impact of theoretical developments for GNNs. In the field of analyzing and learning from data on graphs, GNNs have become an essential tool. With promising applications in different domains such as chemistry, physics, social sciences, knowledge graphs, recommendation, and neuroscience, how to study and build more powerful GNNs is a hot topic.”

Transformer-based models are now widely used in NLP, but we still do not understand a lot about their inner workings. This paper describes what is known to date about the famous BERT model (Devlin et al. 2019), synthesizing over 40 analysis studies. We also provide an overview of the proposed modifications to the model and its training regime. We then outline the directions for further research.

“Sno stores geospatial and tabular data in Git, providing version control at the row and cell level.”

“In the February 2020 issue of Communications of the ACM, Microsoft researcher Patrice Godefroid published a new review article entitled “Fuzzing: Hack, Art, and Science,” which presents an overview of the main automated testing techniques in use today for finding security vulnerabilities in software.”