Machine learning has become a buzzword in the media these days. Recently Science magazine published a cover paper on Human-level concept learning through probabilistic program induction and shortly after Nature magazine devoted its cover story to AlphaGo, an AI program that defeated European Go Championship winner. Late on Tuesday night, Google's DeepMind AI group will play one of the world's best human Go players, Lee Se-dol of South Korea. The game will be live streamed on YouTube, and the stream is embedded at the end of this story. Many are now discussing the potential of artificial intelligence, asking questions such as "Can machines learn like a human?", "Will artificial intelligence surpass human intelligence?", To answer such questions, InfoQ interviewed Prof. Jürgen Schmidhuber, Scientific Director of The Swiss AI Lab IDSIA.

The purpose of this study has been to gauge corporate attitudes toward AI in different regions and different industries. Based on a global survey of 203 senior executives, it finds that, especially in North America, companies in health and life sciences, in retail, in manufacturing and in financial services are actively testing the waters. Amongst this group, AI technologies and applications are in the exploratory phase at around one-third of companies, but another third have moved on to experimentation, and onetenth have begun to utilise AI in limited areas. A small handful (2.5%) have even deployed it widely.

IBM's chief financial officer, Martin Schroeter, made a bold statement on the future of technology on Thursday -- and he got some pushback. "The debate about whether artificial intelligence is real is over," Schroeter said, referring to IBM's cognitive computing platform, Watson. "And we're getting to work to solve real business problems." Katy Huberty, managing director at Morgan Stanley, challenged IBM on the topic during Thursday's earnings conference call. Huberty: "My other question is Watson: From the outside, it seems this business gets a pretty significant share of the press, but not contributing to revenue. Do you have visibility yet to when we can expect an inflection in revenue recognition from Watson? Or should we just not think about this as a contributing factor, or moving the needle in our models over the next couple of years? Schroeter: "....Watson is a silver thread.

Artificial intelligence and deep learning are shaping up as the next big paradigm shift in computing and several major chipmakers are poised to benefit, Mizuho Securities analyst Vijay Rakesh said in a research report Wednesday. "We believe deep learning and AI with parallel processing (are) driving broad industry adoption as the enterprise segment looks to use available, real-time data to learn, predict, and prepare for contingencies better and faster," Rakesh said. Deep learning, machine learning and artificial intelligence are "being broadly adopted in health care, manufacturing, automotive, finance, insurance, banking, and retail." Graphics chipmaker Nvidia (NVDA) is "well-positioned for the next decade" in the market, Rakesh said. He also sees opportunities for Advanced Micro Devices (AMD) and Intel (INTC) to provide field-programmable gate arrays.

Miguel Ferrer, an actor with a long list of credits ranging from "Twin Peaks" to his current role on CBS' "NCIS: Los Angeles," died of cancer on Thursday. A fixture on TV and in movies since the 1980s, Ferrer's reputation as a scene-stealer began with 1987's "RoboCop," where he played Bob Morton, the conniving corporate executive who designed the film's title cyborg. His other landmark role was as FBI agent Albert Rosenfield in David Lynch's landmark series "Twin Peaks," along with its corresponding film, "Fire Walk With Me." Ferrer reprised the role in the upcoming return of the series, which is set to debut in May on Showtime. "Great talent, better man," wrote "Twin Peaks" co-creator Mark Frost on Twitter. "Working & writing for him was a highlight in every part of my life."

We consider the problem of learning a policy for a Markov decision process consistent with data captured on the state-actions pairs followed by the policy. We assume that the policy belongs to a class of parameterized policies which are defined using features associated with the state-action pairs. The features are known a priori, however, only an unknown subset of them could be relevant. The policy parameters that correspond to an observed target policy are recovered using $\ell_1$-regularized logistic regression that best fits the observed state-action samples. We establish bounds on the difference between the average reward of the estimated and the original policy (regret) in terms of the generalization error and the ergodic coefficient of the underlying Markov chain. To that end, we combine sample complexity theory and sensitivity analysis of the stationary distribution of Markov chains. Our analysis suggests that to achieve regret within order $O(\sqrt{\epsilon})$, it suffices to use training sample size on the order of $\Omega(\log n \cdot poly(1/\epsilon))$, where $n$ is the number of the features. We demonstrate the effectiveness of our method on a synthetic robot navigation example.

From Table II and Figure 1 and 2, GRU1 and GRU2 perform almost as well as GRU0 on MNIST pixel-wise generated sequence inputs. While GRU3 does not perform as well for this (constant base) learning rate. Figure 3 shows that reducing the (constant base) learning rate to (0.0001) and below has enabled GRU3 to increase its (test) accuracy performance to 59.6% after 100 epochs, and with a positive slope indicating that it would increase further after more epochs. Note that in this experiment, GRU3 has about 33% of the number of (adaptively computed) parameters compared to GRU0. Thus, there exists a potential tradeoff between the higher accuracy performance and the decrease in the number of parameters.

Confidence intervals are a popular way to visualize and analyze data distributions. Unlike p-values, they can convey information both about statistical significance as well as effect size. However, very little work exists on applying confidence intervals to multivariate data. In this paper we define confidence intervals for multivariate data that extend the one-dimensional definition in a natural way. In our definition every variable is associated with its own confidence interval as usual, but a data vector can be outside of a few of these, and still be considered to be within the confidence area. We analyze the problem and show that the resulting confidence areas retain the good qualities of their one-dimensional counterparts: they are informative and easy to interpret. Furthermore, we show that the problem of finding multivariate confidence intervals is hard, but provide efficient approximate algorithms to solve the problem.

Even for a medical discipline steeped in a tradition of randomized trials, the evidence basis for only a few guidelines is based on randomized trials (Tricoci et al., 2009). In part this is due to continued development of treatments, in part to enormous expense of clinical trials, and in large part to the hundreds of treatments and their nuances involved in real-world, heterogeneous clinical practice. Thus, many therapeutic decisions are based on observational studies. However, comparative treatment effectiveness studies of observational data suffer from two major problems: only partial overlap of treatments and selection bias. Each treatment is to a degree bounded within constraints of indication and appropriateness. Thus, transplantation is constrained by variables such as age, a mitral valve procedure is constrained by presence of mitral valve regurgitation. However, these boundaries overlap widely, and the same patient may be treated differently by different physicians or different hospitals, often without explicit or evident reasons. Thus, a fundamental hurdle in observational studies evaluating comparative effectiveness of treatment options is to address the resulting selection bias or confounding. Naively evaluating differences in outcomes without doing so leads to biased results and flawed scientific conclusions.

In this paper, a methodology is investigated for signal recovery in the presence of non-Gaussian noise. In contrast with regularized minimization approaches often adopted in the literature, in our algorithm the regularization parameter is reliably estimated from the observations. As the posterior density of the unknown parameters is analytically intractable, the estimation problem is derived in a variational Bayesian framework where the goal is to provide a good approximation to the posterior distribution in order to compute posterior mean estimates. Moreover, a majorization technique is employed to circumvent the difficulties raised by the intricate forms of the non-Gaussian likelihood and of the prior density. We demonstrate the potential of the proposed approach through comparisons with state-of-the-art techniques that are specifically tailored to signal recovery in the presence of mixed Poisson-Gaussian noise. Results show that the proposed approach is efficient and achieves performance comparable with other methods where the regularization parameter is manually tuned from the ground truth.