Feature subset selection arises in many high-dimensional applications of statistics, such as compressed sensing and genomics. The $\ell_0$ penalty is ideal for this task, the caveat being it requires the NP-hard combinatorial evaluation of all models. A recent area of considerable interest is to develop efficient algorithms to fit models with a non-convex $\ell_\gamma$ penalty for $\gamma\in (0,1)$, which results in sparser models than the convex $\ell_1$ or lasso penalty, but is harder to fit. We propose an alternative, termed the horseshoe regularization penalty for feature subset selection, and demonstrate its theoretical and computational advantages. The distinguishing feature from existing non-convex optimization approaches is a full probabilistic representation of the penalty as the negative of the logarithm of a suitable prior, which in turn enables efficient expectation-maximization and local linear approximation algorithms for optimization and MCMC for uncertainty quantification. In synthetic and real data, the resulting algorithms provide better statistical performance, and the computation requires a fraction of time of state-of-the-art non-convex solvers.

The problem of finding overlapping communities in networks has gained much attention recently. Optimization-based approaches use non-negative matrix factorization (NMF) or variants, but the global optimum cannot be provably attained in general. Model-based approaches, such as the popular mixed-membership stochastic blockmodel or MMSB (Airoldi et al., 2008), use parameters for each node to specify the overlapping communities, but standard inference techniques cannot guarantee consistency. We link the two approaches, by (a) establishing sufficient conditions for the symmetric NMF optimization to have a unique solution under MMSB, and (b) proposing a computationally efficient algorithm called GeoNMF that is provably optimal and hence consistent for a broad parameter regime. We demonstrate its accuracy on both simulated and real-world datasets.

Editor's note: The following is an interview with University of Illinois professor and text analytics guru Bing Liu, conducted by marketing scientist Kevin Gray, in which Liu concisely outlines the current state of the field. Kevin Gray: I see "text analytics" and "text mining" used in various ways by marketing researchers and often used interchangeably. What do these terms mean to you? Bing Liu: My understanding is that the two terms mean the same thing. People from academia use the term text mining, especially data mining researchers, while text analytics is mainly used in industry. I seldom see academics use the term text analytics.

At first, it was surprising to hear the whir of a robot and see a smiling face roll by. Over the past 40 years, telepresence technologies have gone from being complicated, huge, or non-existent to being so ubiquitous and powerful that they're used everywhere from operating rooms to your phone. So, why not send them to space? That's the argument put forward in an article published today in Science Robotics. As people discuss the best ways to explore other planets, some wonder if it's even necessary to send people to the ground--why not just send a robot avatar instead, and basically Skype with Mars? Dan Lester, one of the authors of the paper, says that this is one idea whose time has finally come.

Cisco today unveiled intent-based networking to deliver security and networking that learns, adapts, and gets smarter over time through the use of artificial intelligence. "The network has been foundational to business and society for the last 30 years, and today I really believe we're redefining the network for the next 30 years," Cisco CEO Chuck Robbins said at an event held today in the Dogpatch neighborhood in San Francisco. "What we believe we can do for our customers is translate their business intent into the network. We can automate policy, we can understand what our customers are trying to do. Think about a manufacturing facility that connects all their assets, and then as they need more capacity then the network dynamically provisions based on what they're trying to achieve in their business."

NASA's Mars Reconnaissance Orbiter has spotted all sorts of strange features on the Martian surface, from deep pits to reptilian-looking craters. But, in a recent observation, the instrument caught a glimpse of something far more familiar. A stunning new image shows a look at the rocky mountainside terrain of Mars' Mount Sharp – and, appearing as a bright blue speck at the center, the Curiosity rover can be seen as it presses onward in its uphill mission. In a recent observation, the instrument caught a glimpse of something familiar. A stunning new image shows a look at the rocky mountainside terrain of Mars' Mount Sharp – and, appearing as a bright blue speck at the center, the Curiosity rover can be seen as it presses onward in its uphill mission Curiosity's top speed is 1.5 inches (3.8 cm) per second.

Uber's discrimination investigation recommends dozens of reforms within their company walls. A sign marks a pick-up point for the Uber car service at LaGuardia Airport in New York on March 15, 2017. Travis Kalanick, the combative and embattled CEO of ride-hailing giant Uber, resigned June 20, 2017 under pressure from investors at a pivotal time for the company. SAN FRANCISCO -- In the tumultuous months leading up to Uber CEO and co-founder Travis Kalanick's resignation, the ride-hailing company lost U.S. market share and saw its brand image tarnished, most notably by a former engineer's blog post blasting the ride-hailing company for its sexist work environment. Among several surveys tracking the company's decline: one based on credit card spending, which found over the past two years, Uber's share of rides has dropped to 75% from 90%, according to TXN Solutions.

According to a new study from Oxford and Yale University researchers, those are the years artificial intelligence is slated to take over each of those tasks. The study relied on survey responses of 352 AI researchers who gave their opinions on when in the future machines would replace humans for various tasks. Language translation could outpace human performance by 2024, responses indicated, and robots may be able to write better high-school-level essays than humans in 2026. Ultimately, the researchers found AI could automate all human tasks by the year 2051 and all human jobs by 2136.

Google hopes its Tensor2Tensor library will help accelerate deep-learning research. Google researchers have created what they call "one model to learn them all" for training AI models in different tasks using multiple types of training data. The researchers and the AI-focused Google Brain Team have packaged up the model along with other tools and modular components in its new Tensor2Tensor library, which they hope will help accelerate deep-learning research. The framework promises to take some of the work out of customizing an environment to enable deep-learning models to work on various tasks. As they note in a new paper called'One model to learn them all', deep learning has had success in speech recognition, image classification and translation, but each model needs to be tuned specifically for the task at hand.

The exploded skull of a man who died in the catastrophic eruption of Mount Vesuvius nearly 2,000 years ago has been pieced together giving scientists a unique opportunity to capture the ancient face using 3D imaging. It is the first real-life reconstruction of the features of a victim of the volcanic disaster who lived in the ill-fated seaside town of Herculaneum. The appearance is that of a typical southern European who may have been wealthy and educated because he was 50 years old when he died - an unusual milestone for the time. He was one of 350 casualties discovered frozen in time, buried under volcanic ash in Herculaneum. This is the first real-life reconstruction of the features of a victim of the volcanic disaster who lived in the ill-fated seaside town of Herculaneum.