NYU Data Science newsletter – August 31, 2016

GROUP CURATION: N/A

To remember events in the order they occur, the brain’s neurons function in a coordinated way that is akin to a symphony, a team of NYU scientists has found. Their findings offer new insights into how we recall information and point to factors that may disrupt certain types of memories.

Join us as we explore the rise of artificial intelligence in six charts including the top investors in AI and the most used AI enterprise solutions.

[University of Michigan] researchers developed LobbyBack: a new system of natural language processing analysis that takes clusters of bills and identifies common language, attempting to reconstruct those sentences into something that might resemble the original model legislation.

“This is about trying to figure out what the law looked like in the original document and how laws propagate from these organizations to state legislatures,” [Eytan] Adar said. “There are two reasons why it’s exciting: First, this is an area where that kind of influence has a huge effect. Because it’s not transparent, it’s problematic. It’s also a hard, interesting problem that exists in areas ranging from meme diffusion to plagiarism detection.”

When economic data appear in the media, they are generally discussed as national statistics; for instance, a given number of jobs were added across the country or the economy as a whole grew by a certain percentage during the past quarter. Those data can yield useful information, but they can also mask important regional variations and trends, argues Erik Hurst, an economist at the University of Chicago’s Booth School of Business. … In some ways, I’m [Hurst] perceived as the anti-John Haltiwanger when it comes to entrepreneurship. He and I are often on the same panels, and he’s definitely seen as the glass half full guy and I’m definitely the glass half empty guy. But I think we both are right. His line is there’s a huge amount of job growth that comes from these new small entrepreneurial businesses and that’s very important to the dynamics of the U.S. economy, and I agree 100 percent. My line has always been essentially that most small businesses simply don’t grow. But that doesn’t mean those statements are necessarily inconsistent with each other.

We’re excited to announce a new research partnership with the Radiotherapy Department at University College London Hospitals NHS Foundation Trust, which provides world-leading cancer treatment.
… with clinicians in UCLH’s world-leading radiotherapy team we are exploring whether machine learning methods could reduce the amount of time it takes to plan radiotherapy treatment for such cancers.

We introduce an intuitive model that describes both the emergence of community structure and the evolution of the internal structure of communities in growing social networks. The model comprises two complementary mechanisms: One mechanism accounts for the evolution of the internal link structure of a single community, and the second mechanism coordinates the growth of multiple overlapping communities. The first mechanism is based on the assumption that each node establishes links with its neighbors and introduces new nodes to the community at different rates. We demonstrate that this simple mechanism gives rise to an effective maximal degree within communities. This observation is related to the anthropological theory known as Dunbar’s number, i.e., the empirical observation of a maximal number of ties which an average individual can sustain within its social groups. The second mechanism is based on a recently proposed generalization of preferential attachment to community structure, appropriately called structural preferential attachment (SPA). The combination of these two mechanisms into a single model (SPA+) allows us to reproduce a number of the global statistics of real networks: The distribution of community sizes, of node memberships, and of degrees. The SPA+ model also predicts (a) three qualitative regimes for the degree distribution within overlapping communities and (b) strong correlations between the number of communities to which a node belongs and its number of connections within each community. We present empirical evidence that support our findings in real complex networks.

Michelle Zhou spent over a decade and a half at I.B.M. Research and I.B.M. Watson Group before leaving to become a co-founder of Juji, a sentiment-analysis startup. An expert in a field where artificial intelligence and human-computer interaction intersect, Zhou breaks down A.I. into three stages. The first is recognition intelligence, in which algorithms running on ever more powerful computers can recognize patterns and glean topics from blocks of text, or perhaps even derive the meaning of a whole document from a few sentences. The second stage is cognitive intelligence, in which machines can go beyond pattern recognition and start making inferences from data. The third stage will be reached only when we can create virtual human beings, who can think, act, and behave as humans do.

We are a long way from creating virtual human beings.

“A few hours ago, 50 months after Elsevier submitted a patent application for an “Online peer review system and method” the patent was awarded to the company.”

“HOW or IF Elsevier is currently utilizing or will utilize the technology/method receiving patent is unknown.”

Deploying AI will require a kind of reboot in the way companies think about privacy and security. AI is fueled by data. The more the machine learns about you, the better it can predict your needs and act on your behalf. But as data becomes the currency of our digital lives, companies must ensure the privacy and security of customer information. And, there is no trust without transparency – companies must give customers clarity on how their personal data is used. It turns out that the capability of AI to detect and remedy security breaches plays a critical role in protecting user privacy and building trust.

More on the Future of AI:

The Hype—and Hope—of Artificial Intelligence (August 26, The New Yorker, Om Malik)

What’s the future of Artificial Intelligence? (August 22, Raconteur Media, UK)

UC Berkeley launches Center for Human-Compatible Artificial Intelligence (August 29, University of California-Berkeley, Berkeley News)

G.E., the 124-Year-Old Software Start-Up (August 27, The New York Times)

Children quickly learn to predict what will happen if they turn a cup filled with juice upside down. Robots, on the other hand, don’t have a clue.

Researchers at the Allen Institute for Artificial Intelligence (Ai2) in Seattle have developed a computer program that shows how machines determine how the objects captured by a camera will most likely behave. This could help make robots and other machines less prone to error, and might help self-driving cars navigate unfamiliar scenes more safely.

“This week in Chicago, the Array of Things team begins the first phase of the groundbreaking urban sensing project, installing the first of an eventual 500 nodes on city streets. By measuring data on air quality, climate, traffic and other urban features, these pilot nodes kick off an innovative partnership between the University of Chicago, Argonne National Laboratory and the city of Chicago to better understand, serve and improve cities.”

Also in IoT:

IoT Accelerators Hope to Solve Time-to-Value Challenge (August 29, Research Triangle Institute, RTInsights, Sangeeta Deogawanka)

End-to-End IoT Security Simplified (August 25, EE Times, Rich Quinnell)

Is smart dust the IoT vector of the future? (August 20, ReadWrite, Cate Lawrence)

Tech Billionaire’s Data Startup C3 IoT Raises $70 Million (September 01, Bloomberg, Eric Newcomer)

We are thrilled to announce that Arthur Spirling, the Interim Deputy Director at the Center for Data Science, has been awarded a $750,000 grant from the National Science Foundation for a project titled, “Computational and Historical Resources on Nations and Organizations for the Social Sciences (CHRONOS).” Spirling, along with colleagues and collaborators from Columbia University, will be developing tools to analyze millions of declassified U.S. government records, as a way of studying American foreign relations.

Maria Lavin explains the layout and philosophy of our new academic space

Data is Plural is the newsletter of interesting data sets published by computational journalist Jeremy Singer-Vine.

Earlier this year, we released a TensorFlow implementation of a state-of-the-art image classification model known as Inception-V3. This code allowed users to train the model on the ImageNet classification dataset via synchronized gradient descent, using either a single local machine or a cluster of machines. The Inception-V3 model was built on an experimental TensorFlow library called TF-Slim, a lightweight package for defining, training and evaluating models in TensorFlow. The TF-Slim library provides common abstractions which enable users to define models quickly and concisely, while keeping the model architecture transparent and its hyperparameters explicit.

Since that release, TF-Slim has grown substantially, with many types of layers, loss functions, and evaluation metrics added, along with handy routines for training and evaluating models.

It turns out that over the past two years, deep learning has totally rewritten our approach to machine translation. Deep learning researchers who know almost nothing about language translation are throwing together relatively simple machine learning solutions that are beating the best expert-built language translation systems in the world.

The technology behind this breakthrough is called sequence-to-sequence learning. It’s very powerful technique that be used to solve many kinds problems. After we see how it is used for translation, we’ll also learn how the exact same algorithm can be used to write AI chat bots and describe pictures.

“In this post, we’ll share how deep learning research usually proceeds, describe the infrastructure choices we’ve made to support it, and open-source kubernetes-ec2-autoscaler, a batch-optimized scaling manager for Kubernetes.”

PaddlePaddle (PArallel Distributed Deep LEarning) is an easy-to-use, efficient, flexible and scalable deep learning platform, which is originally developed by Baidu scientists and engineers for the purpose of applying deep learning to many products at Baidu.

Right now, I think that the most exciting industrial applications of optimization are those that synthesize machine learning and optimization in order to obtain optimal personalization at scale.

Here, I’ll talk about a more concrete use case of this synthesis that you might see at a company.