NYU Data Science newsletter – December 4, 2015

GROUP CURATION: N/A

Scientific publishers are forging links with an organization that wants scientists to scribble comments over online research papers.

… Inputs enter the first layer of sigmoid functions, which spits out results that can be combined before being fed into a second layer of sigmoid functions, and so on. This web of resulting functions constitutes the “network” in a neural network. A “deep” one has many layers.

Decades ago, researchers proved that these networks are universal, meaning that they can generate all possible functions. Other researchers later proved a number of theoretical results about the unique correspondence between a network and the function it generates. But these results assume networks that can have extremely large numbers of layers and of function nodes within each layer. In practice, neural networks use anywhere between two and two dozen layers.* Because of this limitation, none of the classical results come close to explaining why neural networks and deep learning work as spectacularly well as they do.

Computer data has been depicted as microscopic magnetic smudges, electric charges and even Lilliputian patterns of dots that reflect laser beams. It may ultimately move into the fabric of life itself — encoded in the organic molecules that are strung together like pearls to form strands of DNA.

In two recent experiments, a team of computer scientists at the University of Washington and Microsoft, and a separate group at the University of Illinois, have shown that DNA molecules can be the basis for an archival storage system potentially capable of storing all of the world’s digital information in roughly nine liters of solution, about the amount of liquid in a case of wine.

We have implemented the FRBNY DSGE model in a free and open-source language called Julia. The code is posted here on GitHub, a public repository hosting service. This effort is the result of a collaboration between New York Fed staff and folks from the QuantEcon project, whose aim is to coordinate development of high performance open-source code for quantitative economic modeling.

You may wonder why we wrote our code, which was originally in MATLAB and made available here, in Julia. MATLAB is a widely used, mature programming language that has served our purposes very well for many years. However, Julia has two main advantages from our perspective. First, as free software, Julia is more accessible to users from academic institutions or organizations without the resources for purchasing a license. Now anyone, from Kathmandu to Timbuktu, can run our code at no cost. Second, as the models that we use for forecasting and policy analysis grow more complicated, we need a language that can perform computations at a high speed.

Data science faces many challenges in the traditional academic setting. At the same time, many research fields are becoming increasingly dependent on data science tools and techniques. A key element in tackling these challenges is the education of a new generation of researchers that are fluent in both their research domain and in data science methodologies. In this post, we discuss an immersive approach to training in data science, the University of Washington eScience Institute’s inaugural Data Science for Social Good (DSSG) program.

It was only a few years ago that driverless cars seemed beyond the realms of possibility and now it’s likely that people will be able to watch them race, with Formula E planning to have them compete at up to 186mph.

The all-electric racing series is working with investment firm Kinetik to create the world’s first driverless car racing series. Dubbed Roborace, the races will run ahead of each scheduled Formula E event from next season.

Social media use is a global phenomenon, with almost two billion people worldwide regularly using these websites. As Internet access around the world increases, so will the number of social media users. Neuroscientists can capitalize on the ubiquity of social media use to gain novel insights about social cognitive processes and the neural systems that support them. This review outlines social motives that drive people to use social media, proposes neural systems supporting social media use, and describes approaches neuroscientists can use to conduct research with social media. We close by noting important directions and ethical considerations of future research with social media.

A number of recent studies have demonstrated how Twitter can be used to anticipate presidential elections, flu outbreaks or box-office results.

So could an analysis of Twitter data also help us predict changes in stock prices?

[Based on the paper Walking Down Wall Street with a Table: A Survey of Stock Market Predictions Using the Web by Nardo, Petracco-Giudici, Naltsidis.]