We give a complete characterization of the complexity of best-arm identification in one-parameter bandit problems. We prove a new, tight lower bound on the sample complexity. We propose the `Track-and-Stop' strategy, which we prove to be asymptotically optimal. It consists in a new sampling rule (which tracks the optimal proportions of arm draws highlighted by the lower bound) and in a stopping rule named after Chernoff, for which we give a new analysis.

Black-box alpha (BB-$\alpha$) is a new approximate inference method based on the minimization of $\alpha$-divergences. BB-$\alpha$ scales to large datasets because it can be implemented using stochastic gradient descent. BB-$\alpha$ can be applied to complex probabilistic models with little effort since it only requires as input the likelihood function and its gradients. These gradients can be easily obtained using automatic differentiation. By changing the divergence parameter $\alpha$, the method is able to interpolate between variational Bayes (VB) ($\alpha \rightarrow 0$) and an algorithm similar to expectation propagation (EP) ($\alpha = 1$). Experiments on probit regression and neural network regression and classification problems show that BB-$\alpha$ with non-standard settings of $\alpha$, such as $\alpha = 0.5$, usually produces better predictions than with $\alpha \rightarrow 0$ (VB) or $\alpha = 1$ (EP).

This paper considers the recovery of a rank $r$ positive semidefinite matrix $X X^T\in\mathbb{R}^{n\times n}$ from $m$ scalar measurements of the form $y_i := a_i^T X X^T a_i$ (i.e., quadratic measurements of $X$). Such problems arise in a variety of applications, including covariance sketching of high-dimensional data streams, quadratic regression, quantum state tomography, among others. A natural approach to this problem is to minimize the loss function $f(U) = \sum_i (y_i - a_i^TUU^Ta_i)^2$ which has an entire manifold of solutions given by $\{XO\}_{O\in\mathcal{O}_r}$ where $\mathcal{O}_r$ is the orthogonal group of $r\times r$ orthogonal matrices; this is {\it non-convex} in the $n\times r$ matrix $U$, but methods like gradient descent are simple and easy to implement (as compared to semidefinite relaxation approaches). In this paper we show that once we have $m \geq C nr \log^2(n)$ samples from isotropic gaussian $a_i$, with high probability {\em (a)} this function admits a dimension-independent region of {\em local strong convexity} on lines perpendicular to the solution manifold, and {\em (b)} with an additional polynomial factor of $r$ samples, a simple spectral initialization will land within the region of convexity with high probability. Together, this implies that gradient descent with initialization (but no re-sampling) will converge linearly to the correct $X$, up to an orthogonal transformation. We believe that this general technique (local convexity reachable by spectral initialization) should prove applicable to a broader class of nonconvex optimization problems.

The satisfiability problem for SPARQL patterns is undecidable in general, since the expressive power of SPARQL 1.0 is comparable with that of the relational algebra. The goal of this paper is to delineate the boundary of decidability of satisfiability in terms of the constraints allowed in filter conditions. The classes of constraints considered are bound-constraints, negated bound-constraints, equalities, nonequalities, constant-equalities, and constant-nonequalities. The main result of the paper can be summarized by saying that, as soon as inconsistent filter conditions can be formed, satisfiability is undecidable. The key insight in each case is to find a way to emulate the set difference operation. Undecidability can then be obtained from a known undecidability result for the algebra of binary relations with union, composition, and set difference. When no inconsistent filter conditions can be formed, satisfiability is efficiently decidable by simple checks on bound variables and on the use of literals. The paper also points out that satisfiability for the so-called `well-designed' patterns can be decided by a check on bound variables and a check for inconsistent filter conditions.

Donald Trump can buy himself many things, but he will never be able to officially build a 10,000-block-high statue of himself in Minecraft. And no, don't expect to mine Moria as part of a Lord of the Rings server. Microsoft and its Mojang subsidiary said Tuesday that they will begin blocking corporations and politicians from using Minecraft to promote their own agendas, including the sale of products, movies, or political views. "We want to empower our community to make money from their creativity, but we're not happy when the selling of an unrelated product becomes the purpose of a Minecraft mod or server," Mojang wrote in a blog post. The new additions are now part of Mojang's Commercial Usage Guidelines.

This year's Computex Taipei 2016 will mark a turning point for Asia's biggest tradeshow, as Taiwan begins asserting itself as being not only the center of the global ICT supply chain but also as a key partner for building the global technology ecosystem and driving innovation. This new positioning for the show will be highlighted in new exhibits including InnoVEX - a startup village - and iSTyle, which will feature a collection of Apple MFi certified products and accessories. A packed schedule of keynotes, panels, forums, and demos provide a rare opportunity for startups to connect with international VCs, angel investors, potential partners, and future customers in a single venue. InnoVEX events will focus on startup technology and entrepreneurial issues, from securing funding and partnerships to building and managing company growth. Participating companies will showcase the latest developments in peripherals, accessories and software for MFi certified products including cables, chargers, and connectors.

AI now plays a pivotal role in so many industries, it simply cannot be ignored -- those who do will simply become charmingly anachronistic. Join this essential VB Live event to understand the current AI landscape and how those who are winning at it, are winning big. Ever since iPhones came equipped with the soothing tones of Siri, artificial intelligence entered the mainstream. Whether answering basic questions or finding information on local architecture, artificial intelligence (AI) simulates human intelligence such as visual perception, speech recognition, decision making and even language translation. AI is revolutionizing products and services throughout the business world.

R. Lee Coulter, one of the organizers of the 2016 World BPO/ITO Forum, is senior vice president of St. Louis-based Ascension Health, the nation's largest nonprofit health system, and CEO of its shared services subsidiary, the Ascension Ministry Service Center. His focus on shared services, business process outsourcing, or BPO, and technology spans a 30-year career in leadership positions at companies, including General Electric, AON and Kraft Foods. Here, he talks to SearchCIO senior executive editor Linda Tucci about the current state of artificial intelligence (AI) applications in outsourcing and what's on the horizon. This interview has been condensed and edited. Lee Coulter: AI is not being applied to outsourcing today, in so far as we're talking about the shared services and outsourcing industry.

Especially when those ideas stand in opposition to technological progress. I'm used to hearing, and quickly dismissing, fears related to our ongoing developments, especially when they come without factual basis or understanding of the underlying societal and technological aspects implicated. So keep all of that in mind when I tell you that I'm terrified of the future. Because I'm about to be fired and replaced by a robot. Yes, I'm afraid I've fallen off the bandwagon once again, knocked my head, and started to agree with some of the pundits you're hearing in the press. While I sense most people have no inkling of the huge tidal wave of change that's coming towards us, some of its effects are already starting to be felt.

Cooney goes on, "With Fun LoL DARPA is looking for information about mathematical frameworks, architectures, and methods that would help answer questions such as: What are the number of examples necessary for training to achieve a given accuracy performance? What are important trade-offs and their implications? How close is the expected achievable performance of a learning algorithm compared to what can be achieved at the limit? What are the effects of noise and error in the training data? What are the potential gains possible due to the statistical structure of the model generating the data?"