Linear machine learning models can be seen as providing two outputs: predictions and weight maps. The latter shows the relative contribution of the individual features to the model and has been heavily used in the neuroimaging community to infer conclusions about brain structure/function. There has however been a recent debate on whether weight maps can provide information about the neural signals leading to a significant classification/regression model [1]-[3]. The authors of [1] indeed suggest that weight maps provide a poor recovery of the input neural signal and lead to false positives. They further demonstrate that the amplitude of the weight does not reflect the amplitude of the signal difference in a feature. However, their examples are specific cases with low signalto-noise ratio (SNR). Here, we investigate the recovery of two widespread techniques, namely SVM [4] and sparse MKL [5] when varying the SNR, as well as the distribution of simulated neural signals.

We present powerful new analysis techniques to constrain effective field theories at the LHC. By leveraging the structure of particle physics processes, we extract extra information from Monte-Carlo simulations, which can be used to train neural network models that estimate the likelihood ratio. These methods scale well to processes with many observables and theory parameters, do not require any approximations of the parton shower or detector response, and can be evaluated in microseconds. We show that they allow us to put significantly stronger bounds on dimension-six operators than existing methods, demonstrating their potential to improve the precision of the LHC legacy constraints.

Dialogue management (DM) decides the next action of a dialogue system according to the current dialogue state, and thus plays a central role in task-oriented dialogue systems. Since dialogue management requires to have access to not only local utterances, but also the global semantics of the entire dialogue session, modeling the long-range history information is a critical issue. To this end, we propose a novel Memory-Augmented Dialogue management model (MAD) which employs a memory controller and two additional memory structures, i.e., a slot-value memory and an external memory. The slot-value memory tracks the dialogue state by memorizing and updating the values of semantic slots (for instance, cuisine, price, and location), and the external memory augments the representation of hidden states of traditional recurrent neural networks through storing more context information. To update the dialogue state efficiently, we also propose slot-level attention on user utterances to extract specific semantic information for each slot. Experiments show that our model can obtain state-of-the-art performance and outperforms existing baselines.

In this work, we propose a subspace-based algorithm for DOA estimation which iteratively reduces the disturbance factors of the estimated data covariance matrix and incorporates prior knowledge which is gradually obtained on line. An analysis of the MSE of the reshaped data covariance matrix is carried out along with comparisons between computational complexities of the proposed and existing algorithms. Simulations focusing on closely-spaced sources, where they are uncorrelated and correlated, illustrate the improvements achieved.

Language recognition system is typically trained directly to optimize classification error on the target language labels, without using the external, or meta-information in the estimation of the model parameters. However labels are not independent of each other, there is a dependency enforced by, for example, the language family, which affects negatively on classification. The other external information sources (e.g. audio encoding, telephony or video speech) can also decrease classification accuracy. In this paper, we attempt to solve these issues by constructing a deep hierarchical neural network, where different levels of meta-information are encapsulated by attentive prediction units and also embedded into the training progress. The proposed method learns auxiliary tasks to obtain robust internal representation and to construct a variant of attentive units within the hierarchical model. The final result is the structural prediction of the target language and a closely related language family. The algorithm reflects a "staircase" way of learning in both its architecture and training, advancing from the fundamental audio encoding to the language family level and finally to the target language level. This process not only improves generalization but also tackles the issues of imbalanced class priors and channel variability in the deep neural network model. Our experimental findings show that the proposed architecture outperforms the state-of-the-art i-vector approaches on both small and big language corpora by a significant margin.

Deep neural networks (DNNs) have achieved great success in solving several longstanding tasks with near human-level intelligence, e.g., the ancient game of Go, image classification, and natural language processing. As a result, many potential applications are envisaged. However, major concerns have been raised about the readiness of applying this technique to safety-and security-critical systems, where faulty behaviour carries the risk of endangering human lives or potential damage to business. To address these concerns, similar to product development in avionics and automotive industries, a (safety or security) critical system implemented with DNNs, or comprising DNNs components, needs to be thoroughly tested and certified. The software industry relies on testing as a primary means to provide stakeholders with information about the quality of the software product or service under test [1].

The convergence of artificial intelligence (AI), big data, automation and the internet of things (IoT) already has widespread implications on the way we design, make and maintain things. These transformative technologies collectively are the drivers of a "fourth industrial revolution". Previous seismic shifts in industrialisation were brought about by the advent of steam, electricity and digital technology. Today, it is data-driven, autonomous and self-learning technologies which are driving the rapid changes we are seeing across many sectors of business and industry. Of course, information has always been the lifeblood of engineering and manufacturing.

Film director Michael Anderson is seen in this undated photo. LONDON – British director Michael Anderson, whose films included war epic "The Dam Busters" and sci-fi classic "Logan's Run," has died at age 98. Anderson's family said Sunday that he died of heart disease April 25 in Canada, at his home on the Sunshine Coast of British Columbia. Born into a theatrical family in London in 1920, Anderson served in the army during World War II and made his feature debut in 1949 with "Private Angelo," co-directed by Peter Ustinov. His 1955 adventure "The Dam Busters" told the story of a daring wartime bombing raid on Germany's industrial heartland. Its visual flair and stirring score helped make it one of Britain's best-loved war films, and its thrilling climax helped inspire the attack on the Death Star in the first "Star Wars" movie.

At the Geneva International Motor Show yesterday, next to the exhibition halls showing off car manufacturers' latest creations, industry experts and UN representatives gathered to discuss how they will fast-forward the automotive industry -- and the world -- into the future. The Symposium on the Future Networked Car (FNC-2018), convened by the International Telecommunications Union (ITU) and the UN Economic Commission for Europe (UNECE), revealed how the automotive industry has been leveraging recent advances in information and communication technologies (ICT) to make transport systems safer, greener, and more intelligent. Participants highlighted the opportunities to be seized and challenges to be overcome for success. Curtis Hay, Technical Fellow at General Motors, described the recently launched Cadillac Super Cruise, which provides a hands-free driving experience. "We need more standards, and the worldwide use of harmonized standards. Connected cars need a lot of power so new standards for 5G can be the basis for connected vehicle standards"–Christoph Nolte, DEKRA A driver just needs to drive the car into a highway lane and, at the right time, push a button and let go, explained Hays, adding that the car performs breaking and collision avoidance.

If I see a robot coming to kill me, I'll just look for the off-switch or how I can unplug it". I'm ever so slightly paraphrasing Pete Trainor, one of yesterday's speakers at Wired Health 2018, but this was the jist of what he said to me over a coffee as we discussed AI, robotics and its place in the the world. And if I can draw a theme from yesterday's excellent Wired Health event held at the very impressive Francis Crick Institute, it's that AI, robotics and digital technology are not here to replace people in healthcare provision. Rather, they are here to augment and "scale-up" the amount that a healthcare professional can do. This was a theme which was alighted on by a number of speakers.