NYU Data Science newsletter – June 14, 2016

GROUP CURATION: N/A

List of Accepted Papers

Earlier this year, Google’s DeepMind taught a computer program to play a wide variety of Atari video games at a superhuman level in a matter of hours. The program was given no background knowledge. It learnt every game from scratch.

Stuart [Russell] observed that “If your newborn baby did that you would think it was possessed”.

Should we be impressed or distressed by such progress?

Ryan Eustice and Ed Olson were into driverless cars before they were cool.

While automakers are searching for partners and making acquisitions in Silicon Valley, this pair of University of Michigan professors are an extremely valuable resource here in the Motor City’s backyard.

Way back in 2007, an eon in the world of technology, they were on competing teams in the DARPA Grand Challenge, a competition of autonomous vehicles sponsored by the research arm of the U.S. Department of Defense.

That is exactly what Forrester wants to find out – is there something behind the AI and Cognitive Computing hype? What my research directors ask, “Is there a there there?”

AI and Cognitive Computing have captured the imagination and interest of organization large and small but does anyone really know how to bring this new capability in and get value from it? Will AI and Cognitive really change businesses and consumer experiences? And the bigger question – WHEN will this happen?

Each year, tens of millions of people around the world are diagnosed with diabetes, high blood pressure, or depression. You’d expect that by now, doctors would have settled on a few standard ways to treat these diseases. But you’d be wrong.

A new analysis of common treatments for these three conditions, using a database of 250 million patients’ records from four countries, has found that at least one in 10 patients received a course of drugs that no other patient with the same condition did. In other words, more than 25 million people were essentially being treated by guesswork.

“You’re in a room where treaties are negotiated, and today it’s covered in sticky notes,” said DJ Patil, PhD, chief data scientist in the Office of Science and Technology Policy, as he welcomed guests on behalf of President Obama to the White House last Thursday. An eclectic mix of health-care stakeholders were there for a special workshop, co-hosted by Stanford Medicine X and the Office of Science and Technology Policy, that was focused on a core principle of the President’s Precision Medicine Initiative — engaging participants as partners in research.

Neuroscience has focused on the detailed implementation of computation, studying neural codes, dynamics and circuits. In machine learning, however, artificial neural networks tend to eschew precisely designed codes, dynamics or circuits in favor of brute force optimization of a cost function, often using simple and relatively uniform initial architectures. Two recent developments have emerged within machine learning that create an opportunity to connect these seemingly divergent perspectives. First, structured architectures are used, including dedicated systems for attention, recursion and various forms of short- and long-term memory storage. Second, cost functions and training procedures have become more complex and are varied across layers and over time. Here we think about the brain in terms of these ideas. We hypothesize that (1) the brain optimizes cost functions, (2) these cost functions are diverse and differ across brain locations and over development, and (3) optimization operates within a pre-structured architecture matched to the computational problems posed by behavior. Such a heterogeneously optimized system, enabled by a series of interacting cost functions, serves to make learning data-efficient and precisely targeted to the needs of the organism. We suggest directions by which neuroscience could seek to refine and test these hypotheses.

Google CEO Sundar Pichai talks with The Verge’s Walt Mossberg about why Google thinks it can beat Apple, Amazon and Microsoft to make artificial intelligence easier and more helpful for every consumer. He says Google wants to offer smarter privacy controls, so that users might save certain types of conversations forever and wipe others off the record. Plus: Pichai adds that the company is investing more into its flagship Nexus devices for the Android operating system, but will not make its own smartphone without an OEM partner.

The U.S. government has made huge strides in its open data practices over the last few years. Since it launched in 2009, data.gov has become a crucial source for everything from climate and agricultural data to Department of Education records. For the most part, this new era of data disclosure didn’t happen because Congress passed new laws; it happened through presidential orders and procedural improvements in the Executive Branch.

Unfortunately, it might be just as easy for future administrations to roll back the current open data program. That’s why EFF supports a bill that would mandate public access to government data and urges Congress to pass it.

The Seed Grant Selection Committee is thrilled to announce that, after a rigorous evaluation process involving multiple referee reports, the following grant proposals were chosen for funding by the Moore-Sloan Data Science Environment.

Brian Parker and Christine Vogel: “Statistics Meets Transcriptomics: Time-Series Responses of Post-Transcriptional Regulation By Families of Conserved RNA Structures”

Ralph Grishman and Alastair Smith: “Health and Death of Political Leaders”

Jonathan Winawer and Heiko Müller: “The Standard Cortical Observer”

Florian Knoll and Carlos Fernandez-Granda: “Estimation of Multiple Tissue Compartments from Magnetic-Resonance-Fingerprinting Data”

Preeti Raghavan and Aaditya Rangan: “Determining Treatment Algorithms for Patient Subgroups in Stroke Rehabilitation”

Thomas Kirchner and Kyunghyun Cho: “Image-based Community Asset and Risk Factor Surveillance System using Deep Learning”

Nathaniel Beck and David Sontag: “Applying Machine Learning Methods to Integrated Time Series”

We’re teaming up with New York University and its pioneering deep learning team to start a research collaboration at our new auto tech office in New Jersey.

NYU’s researchers will work with NVIDIA scientists and engineers to create groundbreaking autonomous driving technology. The collaboration between NVIDIA, a leader in deep learning infrastructure and tools, and NYU’s world-class deep-learning faculty will accelerate the development of autonomous vehicles.

At LinkedIn, we have built Voices, a text analytics platform that provides easy access to member feedback about our website and key products. Voices aggregates unstructured text across both internal (e.g. LinkedIn posts, customer support cases, NPS survey results) and external (e.g. social media, such as Facebook and Twitter, news, forums, and blogs) data sources. Structured member data and unstructured textual data from various channels are ingested into HDFS and passed through a suite of text mining functions. This allows Voices to surface relevant insights by various dimensions, such as value proposition, product, sentiment, trending insights, and many other use cases.

We aggregate internal data sources and purchase external data from vendors, who pull relevant information from publicly-available data on social platforms and online news, blogs, and forums. Additional data attributes (e.g. geography, sentiment, and audience segment) enable deep dives into business domains. Voices also includes reviews for major LinkedIn apps from the Apple App Store and Google Play.

To kick off this series on communicating data science, I interview Tyler [Byers] about how he uses his skills in data visualization and effective reporting to collaborate and influence in his career. His advice to those who are talented at rising to the top of Kaggle’s leaderboard, but need help finding their voice when it comes to communicating the insights in their ensemble? Read extensively outside of your domain and listen to stand-up comedy!

The goals of this guide are to provide some instruction on the best way to share data to avoid the most common pitfalls and sources of delay in the transition from data collection to data analysis. The Leek group works with a large number of collaborators and the number one source of variation in the speed to results is the status of the data when they arrive at the Leek group. Based on my conversations with other statisticians this is true nearly universally.