NYU Data Science newsletter – May 17, 2016

GROUP CURATION: N/A

There are three main ingredients to creating artificial intelligence: hardware (compute and memory), software (or algorithms), and data. We’ve heard a lot of late about deep learning algorithms that are achieving superhuman level performance in various tasks, but what if we changed the hardware?

We get news from many media sources, and also through our friends, online and offline. By the time the news reaches us, it may have been retold in interesting ways, which so far have typically not been quantified. Normally it would be difficult to tell how the information that reaches us differs from its original source, because the sharing of the information is dispersed, or the situation itself is evolving. However, in a few cases, the source is better-defined, for example, when a public entity issues a press release.

In a recent study, we collected a sample of press releases by the U.S. Federal Open Market Committee, published speeches by President Barack Obama, as well as press releases from several tech companies and universities. We then gathered de-identified Facebook data, analyzed in aggregate, on shares of the articles covering the source and the corresponding comments, as shown in the diagram above.

A Seattle startup gets on the artificial-intelligence train with work in natural language understanding and by helping developers create voice controls used on devices.

Forget Nets games — ’net games are coming to Barclays Center!

The organizers of an international video-game tournament circuit expect thousands of fans to cram into Brooklyn’s biggest area to watch a high-stakes “Counter Strike” contest in October. Staring at a screen might sound like something you can just do at home, but experts say it is genuinely thrilling to cheer, gasp, and groan alongside other spectators while watching contestants fight to the digital death.

“It’s a shared experience with a massive amount of other human beings,” said Andy Nealen, a professor of game engineering and computer graphics at New York University. “Just to straight up be in the atmosphere, it’s pretty exhilarati­ng.”

… “Everyone should know cybersecurity,” says William Yang, an 18-year-old from the Arkansas School for Mathematics, Sciences and the Arts in Hot Springs, Arkansas, who tells me security needs to be baked into more software. In the meantime, he’s having fun just figuring out the solution to the cyberpuzzle.

Yang is here at New York University’s Brooklyn campus for Cyber Security Awareness Week. The largest event of its kind, CSAW comprises six competitions for high-school and university students and a career fair. At stake is more than $450,000 in college scholarships for the high schoolers and more than $11,000 in cash prizes for university winners.

Heidi Imker, Director of the Research Data Service and Associate Professor at the University of Illinois Library, will lead local participation in a new Data Curation Network funded by the Sloan Foundation. The grant will enable six institutions, including Illinois, to pilot a “network of expertise” model for data curation services.

Mental illness is one of the most pressing public health issues of our time. While counseling and psychotherapy can be effective treatments, our knowledge about how to conduct successful counseling conversations has been limited due to lack of large-scale data with labeled outcomes of the conversations. In this paper, we present a large-scale, quantitative study on the discourse of text-message-based counseling conversations.

Movidius has announced the release of the Fathom Neural Compute Stick—a deep learning acceleration module—as well as Fathom deep learning software framework, both of which will enable neural networks to be moved out of the cloud and deployed natively in end-user devices.

The Fathom Neural Compute Stick, a USB stick embedded neural network accelerator, features the Myriad 2 vision processor and can run fully-trained neural networks at under 1 Watt of power. Targeted at deep learning product developers and researchers, the Fathom Neural Compute Stick accepts networks defined by Caffe or TensorFlow and their accompanying dataset. It then uses the Movidius Fathom Tool to prepare and execute the convolutional neural network on Myriad 2.

When connected to a PC, the Fathom Neural Compute Stick behaves as a neural network profiling and evaluation tool, meaning companies will be able to prototype faster and more efficiently, reducing time to market for products requiring artificial intelligence.

Deep learning is different, and I believe this fervently because I’ve seen the approach deliver record-beating results in practical applications across an amazing variety of different problems. That’s why TensorFlow is so important to me personally. …
It’s also why I was over the moon to see another Google research team release Parsey McParseface!

OPODIS is an open forum for the exchange of state-of-the-art knowledge on distributed computing and distributed computer systems. All aspects of distributed systems are within the scope of OPODIS.

Madrid, Spain Tuesday-Friday, December 13-16.

Deadline for submissions is Monday, August 22.

There are many possible reasons that could explain this problem. There could be a technical explanation — we implemented backpropagation incorrectly — or, we chose a learning rate that was too high, which in turn let to the problem that we were overshooting the local minima of the cost function.

Deep learning algorithms use large amounts of data and the computational power of the GPU to learn information directly from data such as images, signals, and text. NVIDIA DIGITS offers an interactive workflow-based solution for image classification. Deep learning frameworks offer more flexibility with designing and training custom deep neural networks and provide interfaces to common programming language. The NVIDIA Deep Learning SDK offers powerful tools and libraries for the development of deep learning frameworks such as Caffe, CNTK, TensorFlow, Theano, and Torch.

I presented “Building Data Pipelines in Python”, with a focus on the need to bring R&D and Engineering together, and how basic engineering principles can be beneficial even if your job is not all about writing code. After presenting a very similar talk at PyCon Italy, I found the audience in London to be a bit more on the academic side than I initially thought, which was perfect for my engineering rants. After the usual first few minutes of feeling awkward when speaking publicly, I started my discussion on unit testing and asked how many in the audience write unit tests regularly. Random guy from the audience: “What’s a unit test?”. Thank you kind stranger, you lifted my spirit and the rest of the talk was a breeze