NYU Data Science newsletter – November 3, 2015

GROUP CURATION: N/A

With Google Scholar, PubMed, and other free academic databases at their fingertips, scientists may feel they have plenty of resources to trawl through the ever-growing science literature.

But a search engine unveiled on 2 November by the non-profit Allen Institute for Artificial Intelligence (AI2) in Seattle, Washington, is working towards providing something different for its users: an understanding of a paper’s content. “We’re trying to get deep into the papers and be fast and clean and usable,” says Oren Etzioni, chief executive officer of AI2.

During the press conference for the release of the Autopilot, Tesla CEO Elon Musk referred to each Model S owners as an “expert trainer” – meaning that each driver will train the autonomous features of the system to feed the collective network intelligence of the fleet by simply driving the electric vehicle on Autopilot.

He said that the system should improve every day, but that improvements might only become noticeable every week or so by adding up. Just a few weeks after the release, Model S owners are already taking to the Tesla Motors Club forum to describe how the Autopilot is improving…

For more than a decade, Baidu has been the top provider of search for China’s Internet population. Now Chief Executive Officer Robin Li’s top priority has shifted to online-to-offline services and connecting Baidu’s millions of users to businesses around world. But what’s most interesting is that much of the research for Baidu’s future is actually not done in Beijing but in Silicon Valley, right in Google’s backyard. Bloomberg’s Emily Chang gets an exclusive look inside Baidu’s AI lab in Sunnyvale, California.

We present a machine learning-based methodology capable of providing real-time (“nowcast”) and forecast estimates of influenza activity in the US by leveraging data from multiple data sources including: Google searches, Twitter microblogs, nearly real-time hospital visit records, and data from a participatory surveillance system. Our main contribution consists of combining multiple influenza-like illnesses (ILI) activity estimates, generated independently with each data source, into a single prediction of ILI utilizing machine learning ensemble approaches. Our methodology exploits the information in each data source and produces accurate weekly ILI predictions for up to four weeks ahead of the release of CDC’s ILI reports. We evaluate the predictive ability of our ensemble approach during the 2013–2014 (retrospective) and 2014–2015 (live) flu seasons for each of the four weekly time horizons. Our ensemble approach demonstrates several advantages: (1) our ensemble method’s predictions outperform every prediction using each data source independently, (2) our methodology can produce predictions one week ahead of GFT’s real-time estimates with comparable accuracy, and (3) our two and three week forecast estimates have comparable accuracy to real-time predictions using an autoregressive model. Moreover, our results show that considerable insight is gained from incorporating disparate data streams, in the form of social media and crowd sourced data, into influenza predictions in all time horizons.

We are excited to announce Berkeley’s involvement in the National Science Foundation’s Big Data Regional Innovation Hubs program (see news release below). BIDS senior fellow Michael Franklin is a co-investigator on this endeavor, and the Western Hub will be closely aligned with BIDS activities. Look for more information on this exciting program in the near future.

Open sharing of data that are collected with smart devices would empower citizens and create jobs, say Dirk Helbing and Evangelos Pournaras.

The human-computer combination will be a winning combination in digital health, according to Joi Ito, director for the MIT Lab, who spoke at the Partners HealthCare Connected Health Symposium in Boston this week.

“I think there was an announcement recently that Watson is almost finished with med school,” Ito said. “It’s sort of a joke, but sort of true. I can imagine Watson being able to ingest all of the data that traditionally might be given to a student in med school. Now imagine if you had a computer that had all of the knowledge that you needed for med school and if it were available all of the time, maybe there’s an argument to be made that you don’t have to memorize it if it’s available all of the time.”

On August 26th we held our second online discussion on de-identification and anonymization – this time with the formidable data scientist Sara-Jayne Terp at the helm of the discussion. The focus of this discussion was ‘Risks & Mitigations of Releasing Data’ and this blog post is a summary of what was discussed.

I really enjoyed the cheeky blog post by my pal Rob Story.

Like many other data tool creators, I’ve been annoyed by the assorted “Python vs R” click-bait articles and Hacker News posts by folks who in all likelihood might not survive an interview panel with me on it.

The worst part of the superficial “R vs Python” articles is that they’re adding noise where there ought to be more signal about some of the real problems facing the data science community. Let me say some very brief words about my present perspective on this.