We develop an approach to risk minimization and stochastic optimization that provides a convex surrogate for variance, allowing near-optimal and computationally efficient trading between approximation and estimation error. Our approach builds off of techniques for distributionally robust optimization and Owen's empirical likelihood, and we provide a number of finite-sample and asymptotic results characterizing the theoretical performance of the estimator. In particular, we show that our procedure comes with certificates of optimality, achieving (in some scenarios) faster rates of convergence than empirical risk minimization by virtue of automatically balancing bias and variance. We give corroborating empirical evidence showing that in practice, the estimator indeed trades between variance and absolute performance on a training sample, improving out-of-sample (test) performance over standard empirical risk minimization for a number of classification problems.

An undercover sex sting in Ohio picked up a suspect this week, as a Youngstown State student reportedly wanted to have sex with a 15-year-old boy who was, in reality, an undercover officer. Albert Maruna IV, 22, was arrested and charged Tuesday for unlawful sexual conduct with a minor, among other things. Police say Maruna started talking to the fictitious teen boy in early December on an online dating app. According to the report, Maruna told the undercover officer that he "didn't believe in age." The conversations became explicitly sexual in nature, with Maruna sending nude pictures and suggesting the two should get married someday.

The most dramatic cybersecurity story of 2016 came to a quiet conclusion Friday in an Anchorage courtroom, as three young American computer savants pleaded guilty to masterminding an unprecedented botnet--powered by unsecured internet-of-things devices like security cameras and wireless routers--that unleashed sweeping attacks on key internet services around the globe last fall. What drove them wasn't anarchist politics or shadowy ties to a nation-state. It was a hard story to miss last year: In France last September, the telecom provider OVH was hit by a distributed denial-of-service (DDoS) attack a hundred times larger than most of its kind. Then, on a Friday afternoon in October 2016, the internet slowed or stopped for nearly the entire eastern United States, as the tech company Dyn, a key part of the internet's backbone, came under a crippling assault. As the 2016 US presidential election drew near, fears began to mount that the so-called Mirai botnet might be the work of a nation-state practicing for an attack that would cripple the country as voters went to the polls.

Interest in machine learning has exploded over the past decade. You see machine learning in computer science programs, industry conferences, and the Wall Street Journal almost daily. For all the talk about machine learning, many conflate what it can do with what they wish it could do. Fundamentally, machine learning is using algorithms to extract information from raw data and represent it in some type of model. We use this model to infer things about other data we have not yet modeled.

The Tepper School of Business developed the curriculum for the online Master of Science in Business Analytics (MSBA) program from the ground up with this question in mind. In consultation with global business leaders, they determined that the greatest need is for professionals who not only have advanced analytical skills, such as machine learning and optimization, but also the appropriate business knowledge and communication skills to solve complex problems and bring value to industry. Our students develop proficiency in the full range of state-of-the-art business analytics techniques; they also learn how to tell stories through and extract insights from data. Given the Tepper School's view of a curriculum as an organic entity, our faculty continually work in concert to ensure that courses harmonize, even as they are individually updated and modified to ensure learning outcomes for students are always in step with an ever-evolving industry. The flexible online format enables students to continue working while earning their degree and apply what they learn in the classroom to their work environment.

For many years, the American public thought of artificial intelligence, or AI, as some big, incredibly complex computer program that would one day achieve sentience and turn on humanity like a James Cameron conceived summer blockbuster. While movie franchises starring angry AI creations turning on their creators continue to do well at the box office, artificial intelligence has made its way into the modern American household. From algorithms on Amazon that suggest what we might want to read next to Siri and Alexa answering our questions to cars that understand traffic patterns and very soon might be driving us to work, artificial intelligence is rapidly disrupting industries. While there has been a great fear for a number of years that robots will replace workers across industries, this has largely not been born out. While robots powered by AI can often replace workers at some of the most menial, automated tasks in a factory, workers are often needed in more specialized capacities to repair, maintain, and respond to alarms that these robotic workers generate.

University of Maryland and Capital One on Tuesday announced a partnership to develop a workforce pipeline in data analytics, machine learning and cybersecurity. Capital One is investing $3 million through an endowment gift to advance machine learning and the two plan to create an innovation lab where students will apply classroom lessons to real-world problems. "This partnership will not only help attract and retain top faculty and students, but will also propel UMD to national prominence and excellence in these critically important fields," said Mary Ann Rankin, senior vice president and provost at University of Maryland. Just over $2 million of Capital One's gift will be used to endow a faculty chair in the department of computer science. The remaining $900,000 will support research and educational initiatives in machine learning, data analytics and cyber security.

A security robot is shooing away homeless people from outside smart office buildings as it patrols the streets of San Francisco. The non-profit organisation that occupies the office block has been warned by officials they will be fined $1,000 (£750) a day if they continue to use the robot without a permit. The Society for the Prevention of Cruelty to Animals (SPCA) say the security robot - dubbed K9 - was hired to deal with the growing amounts of crime related to homeless people on the sidewalks. The crime-fighting robots rely on cameras, Lidar, thermal-imaging to navigate the streets. Laser scanning can detect changes in an environment, while odour detectors can also detect other changes in the area and monitor air pollution.

IBM researchers have developed a program that can predict the products of organic chemistry reactions.1 Modelled on the latest language translation systems – like Google's artificial neural network – the AI picked the right product 80% of the time despite not having been taught any organic chemistry rules. 'What this tool is trying to do is imitate a top pro chemist in more or less the entire domain of organic chemistry,' says Teodoro Laino, one of the researchers involved in the study at IBM in Zurich, Switzerland. His ambitious goal is shared by other chemists who have been attempting to create a functioning AI chemist since the 1970s, when organic chemist E J Corey kick-started the field by creating a chemical knowledge database. However, making a tool based on chemistry knowledge can be time-consuming; Bartosz Grzybowski's team took 10 years to encode their Chematica retrosynthesis program with 20,000 chemical rules. Moreover, a knowledge-based AI has difficulty tackling reactions that lie outside of its rule set. 'There's a way to learn organic chemistry that's not memorising chemical rules, by just trying to find out the underlying patterns in reactions and trying to rationalise them,' Laino says, explaining the approach that his team took.

Google has massively cut prices for its managed cloud machine learning service just two weeks after AWS released a competing offering at its re:Invent user conference. The company has introduced massive price reductions for its Cloud Machine Learning Engine managed services. For example, customers using basic-tier compute for training a machine learning system will pay 43 percent less than they did earlier this year. Google also offered customers more clarity on what they'll be paying for those jobs. Information of the price reductions was first included in a blog post that appeared briefly yesterday on Google's website, then vanished.