Conservatives are girding for an extended clash on two fronts in the months ahead: one with a possible Clinton administration and one with Republicans who rejected Donald J. Trump. Megyn Kelly's divergent approach at Fox News took a different turn in her exchange with Newt Gingrich and again raised the question of the channel's future. A lot of healthy people are defying predictions by the Affordable Care Act architects and refusing to enroll, throwing off the calculations behind the system. The startling double-digital declines in TV viewership raise questions about whether the football and soccer leagues have reached their peak. Mr. Beatty's "Rules Don't Apply" is the first film he has written, directed and starred in since "Bulworth" in 1998.

HARI SREENIVASAN: More than ever before, a big part of this election campaign has played itself out on social media. No doubt the candidates and their campaigns have tried to take advantage of these platforms. But there's been a much bigger role this year as well for unseen players. You might call it the rise of the bots. Miles O'Brien has the story, part of our weekly reporting about on the Leading Edge of science and technology.

Finding the right piece of data can be a big challenge. With the latest release of Collibra's data governance and data catalog solution, machine learning algorithms help the product learn what types of data you use with the goal of surfacing and recommend new data sources that are appropriate for your job. This use of machine learning is one of the features in Collibra 5.0, the latest release of the company's flagship data governance solution that was formally announced yesterday. The Collibra Catalog is one of applications enabled atop this core platform that hundreds of companies use to keep track of big data sitting in Hadoop, Hive, and other locations. "We have a technology platform that has the capability to keep track of processes around data, the metadata and organizations and roles and who has responsibility for data," says Daniel Sholler, director of product marketing for Collibra.

We describe a novel optimization method for finite sums (such as empirical risk minimization problems) building on the recently introduced SAGA method. Our method achieves an accelerated convergence rate on strongly convex smooth problems. Our method has only one parameter (a step size), and is radically simpler than other accelerated methods for finite sums. Additionally it can be applied when the terms are non-smooth, yielding a method applicable in many areas where operator splitting methods would traditionally be applied.

Signal subspace methods (SSM) are efficient techniques to reduce dimensionality of data and to filter out noise [1]. The fundamental idea under SSM is to project the data on a basis made of two subspaces, one mostly containing the signal and the other the noise. The two subspaces are separated by a thresholding criterion associated with some measures of information. The two most popular methods of signal subspace decomposition are wavelet shrinkage [2] and Principal Component Analysis (PCA) [3]. Both techniques have proved to be quite efficient. However, wavelet decomposition depending on signal statistics is not equally adapted to different data, and requires some knowledge on prior distributions or parameters of signals to efficiently choose the thresholds for shrinkage. A significant advantage of the PCA is its adaptability to data. The separation criterion is based on energy which may be seen as a limitation in some cases as illustrated in the next section. In recent years, sparse coding has attracted significant interest in the field of signal denoising [4].

There are now many publicly available Gaussian process libraries ranging in scale from personal projects to major community tools. We will therefore only consider a relevant subset of the existing libraries. The influential GPML toolbox (Rasmussen and Nickisch, 2010) uses MATLAB. It has been widely forked. A key reference for our particular contribution is the GPy library (GPy, since 2012), which is written primarily using Python and Numeric Python (NumPy). GPy has an intuitive object-oriented interface. Another relevant Gaussian process library is GPstuff (Vanhatalo et al., 2013) which is also a MATLAB library.

In this paper, we propose a Double Thompson Sampling (D-TS) algorithm for dueling bandit problems. As indicated by its name, D-TS selects both the first and the second candidates according to Thompson Sampling. Specifically, D-TS maintains a posterior distribution for the preference matrix, and chooses the pair of arms for comparison by sampling twice from the posterior distribution. This simple algorithm applies to general Copeland dueling bandits, including Condorcet dueling bandits as its special case. For general Copeland dueling bandits, we show that D-TS achieves $O(K^2 \log T)$ regret. For Condorcet dueling bandits, we further simplify the D-TS algorithm and show that the simplified D-TS algorithm achieves $O(K \log T + K^2 \log \log T)$ regret. Simulation results based on both synthetic and real-world data demonstrate the efficiency of the proposed D-TS algorithm.

U .S. military officers watched grainy video feeds at a small operations center in Baghdad on Tuesday as Predator drones tracked and killed three reputed Islamic State leaders -- one after another -- in the offensive on Mosul. The targeted air strikes were due in large part to intelligence extracted from cellphones, computer hard drives, memory cards and hand-written ledgers recovered from battlefields and towns taken from Islamic State fighters. Recently captured intelligence also has proved useful in providing clues to detecting potential terrorist plots, tracking foreign fighters and identifying Islamic State supporters around the globe, U.S. officials said. The largest data trove was recovered when U.S.-backed Syrian rebel forces recaptured Manbij, an Islamic State stronghold in northern Syria, in mid-August. Intelligence agencies recovered more than 120,000 documents, nearly 1,200 devices and more than 20 terabytes of digital information, officials said. Islamic State militants came early in the morning, riding atop trucks that lumbered into this northern Iraqi oil town.

Connected devices, Smart City, home automation, e-health, Big Data ... In recent years, the concepts of communicating objects have multiplied. In reality, they are all one facet of the same upheaval - the Internet of Things. Cars can be driven without a driver, TVs are going online, and heating systems are activated automatically to the arrival of the residents. The Internet is making many processes in daily life easier. The Internet of Things, or IoT, which enable devices to communicate with people or machines, is actually working in many places in our daily life already.

Yoshua Bengio, one of the leading figures behind the rise of deep learning, is launching a Silicon Valley-style startup incubator dedicated to this enormously influential form of artificial intelligence. The incubator, Element AI, will help build companies from AI research that emerges from the University of Montreal, where Bengio is a professor, and nearby McGill University, and he says this is just part of his efforts to develop an "AI ecosystem" in Montreal. Bengio says the Canadian city offers "the biggest concentration in the world" of academic researchers exploring deep learning, the breed of AI that now plays such an important role inside the likes of Google, Facebook, and Microsoft. "Element AI will help entrepreneurs get started in that high-growth area, with a team of experts--and my help--to steer those companies in the right direction," he says. According to Bengio, about 100 researchers are exploring deep learning at the University of Montreal and about 50 others are doing similar work at McGill.