NYU Data Science newsletter – August 15, 2016

GROUP CURATION: N/A

In this episode of the O’Reilly Data Show, I spoke with Jana Eggers, CEO of Nara Logics. Eggers’ involvement with AI dates back to her days as a researcher at the Los Alamos National Laboratory. Most recently she has been helping companies across many industries adopt AI technologies as a way to enable a range of intelligent data applications. [audio, 34:35]

Can machine learning help government agencies track down terrorists? A secretive arm of the business intelligence firm SAP says yes.

Session with François Chollet,
Deep learning researcher at Google, author of Keras

We explored whether use of deep learning to model temporal relations among events in electronic health records (EHRs) would improve model performance in predicting initial diagnosis of heart failure (HF) compared to conventional methods that ignore temporality. … Deep learning models adapted to leverage temporal relations appear to improve performance of models for detection of incident heart failure with a short observation window of 12–18 months. [full text]

Engineers at Fujitsu Laboratories have developed new software that can speed up deep learning projects run over multiple GPUs.

According to Fujitsu Labs officials, tests have found that the software used with 16 and 64 GPUs are 14.7 to 27 times faster than using a single GPU to run deep learning workloads, with increases in learning speeds 46 percent (on 16 GPUs) to 71 percent (on 64 GPUs). This is important given the increasing popularity of deep learning, a subset of machine learning, which is foundational to the development of artificial intelligence (AI).

For Steven Keating, an engineering PhD candidate at MIT, his curiosity for collecting personal data is what helped him discover his own brain tumor, and saved his life.

After opting into an on-campus research project calling for volunteers to have brain scans, the research team found an abnormality in Keating’s brain, but they thought nothing of it.

A few years later, Keating noticed he was occasionally smelling a vinegar scent, went back to analyze the data from the original scan, and saw that the abnormality in his brain was located next to the olfactory system—the smell center of the brain.

Thanks to a sleek new computer chip developed by IBM, we are one step closer to making computers work like the brain.

The neuromorphic chip is made from a phase-change material commonly found in rewritable optical discs (confused? more on this later). Because of this secret sauce, the chip’s components behave strikingly similar to biological neurons: they can scale down to nanometer size and perform complicated computations rapidly with little energy.
ibm-artificial-neuron-chip-3 (1)

What makes them especially amazing is how they “fire.” They integrate previous input history to determine whether or not to activate. They also show a characteristic trait of biological neurons called stochasticity — that is, when given a similar input, the chip always produces a slightly different, unpredictable result. Stochasticity is the basis of population coding, a type of highly efficient computation that relies on groups of neurons working together. This neuronal quirk was previously tough to mimic using artificial materials.

One of the core problems of modern statistics is to approximate difficult-to-compute probability distributions. This problem is especially important in Bayesian statistics, which frames all inference about unknown quantities as a calculation involving the posterior distribution. In this paper, we review variational inference (VI), a method from machine learning that approximates probability distributions through optimization. VI has been used in many applications and tends to be faster than classical methods, such as Markov chain Monte Carlo sampling. The idea behind VI is to first posit a family of distributions and then to find the member of that family which is close to the target. Closeness is measured by Kullback-Leibler divergence. We review the ideas behind mean-field variational inference, discuss the special case of VI applied to exponential family models, present a full example with a Bayesian mixture of Gaussians, and derive a variant that uses stochastic optimization to scale up to massive data. We discuss modern research in VI and highlight important open problems. VI is powerful, but it is not yet well understood. Our hope in writing this paper is to catalyze statistical research on this class of algorithms. [link to full text pdf]

The University of California San Francisco’s Clinical and Translational Science Institute (CTSI) has received an $85 million funding commitment from the National Institutes of Health (NIH) to advance research into precision medicine and personalized care.

The five-year grant will support the development of a precision medicine biobanking effort, as well as investigations into how healthcare organizations can leverage electronic health records to conduct research and tailor treatments to an individual patient’s needs.

In May, the OpenTrialsFDA team (a collaboration between Erick Turner, Dr. Ben Goldacre and the OpenTrials team at Open Knowledge) was selected as a finalist for the Open Science Prize. Working towards a first prototype in early December, OpenTrialsFDA will make the Drug Approval Packages (DAPs) from the FDA website easily accessible and searchable and link these to documents and data related to clinical trials. Other interested parties will also be able to access, search and present this information through the application programming interfaces (APIs) the team will produce.

Since the commercial cannabis industry is continuously expanding at a very high rate, we now have a lot of data to quantify and segment. As users, we sometimes find ourselves swimming in uncharted seas of information and choice.

In this sense, we have thought about the public’s pain points, the constantly-evolving data science and how we, as cannabis enthusiasts, can contribute to this.

That is what ignited Weedster as a project: coming up with a friendly and accessible way of finding the most relevant cannabis strains matching the effects sought at specific times of the day, all available at dispensaries nearby

Armed with effective narrative and backed by aggressive lawmakers, the upstart has steadily landed more federal business and is now shouldering its way into the Army acquisition system.

We’re always looking for smart ways to learn more in less time. This mentality has led our researchers to push the boundaries of the data we can record in our research process. Alongside the qualitative data we gather from interviews, we also pull together behavioral data collected over longer time periods. We might record someone’s steps per day, or the route they take to work. We’re always looking for insights into the patterns of everyday.

However, more data means more time spent managing data. So about four months ago, I pulled together this little helper, an automated assistant that I coded to save our design team a little time. We even gave it a little personality to make living with it more fun.

“The competition is open to all students (2 and 4 year undergraduate and graduate students) who are U.S. citizens. Individual submissions and team submissions (up to three people) will be accepted. All teams must identify a project lead. The project lead is responsible for submitting the application. Submissions must address one of the three themes described in the evaluation criteria below.”

Deadline to apply is Thursday, September 15.

Scikit-Optimize, or skopt, is a simple and efficient library for sequential model-based optimization, accessible to everybody and reusable in various contexts.

Effective machine learning means building expressive models that sift out signal from noise—that simplify the complexity of real-world data, yet accurately intuit and capture its subtle underlying patterns.

Whatever the downstream application, a primary challenge often boils down to this: How do we represent, or even synthesize, complex data in the context of a tractable model?

This challenge is compounded when working in a limited data setting—especially when samples are in the form of richly-structured, high-dimensional observations like natural images, audio waveforms, or gene expression data.

Cue the Variational Autoencoder, a fascinating development in unsupervised machine learning that marries probabilistic Bayesian inference with deep learning.

I’d like to move this into a proper matrix where we can list implementer feedback and support, but for now here’s a basic list of what is new in SVG 2.

Amazon initially launched the Amazon API Gateway last year. The Gateway allows developers to build backend services for mobile, Web, enterprise, and IoT applications. Now, Amazon has introduced Usage Plans for the API Gateway that will allow developers to build and monetize APIs as well as create an ecosystem around such APIs. Usage Plans come in varying levels of access (gold, silver, bronze; individual, professional, enterprise; and many more). The move hopes to keep pace with competitors who already offer similiar options.