Chemotherapeutic response of cancer cells to a given compound is one of the most fundamental information one requires to design anti-cancer drugs. Recent advances in producing large drug screens against cancer cell lines provided an opportunity to apply machine learning methods for this purpose. In addition to cytotoxicity databases, considerable amount of drug-induced gene expression data has also become publicly available. Following this, several methods that exploit omics data were proposed to predict drug activity on cancer cells. However, due to the complexity of cancer drug mechanisms, none of the existing methods are perfect. One possible direction, therefore, is to combine the strengths of both the methods and the databases for improved performance. We demonstrate that integrating a large number of predictions by the proposed method improves the performance for this task. The predictors in the ensemble differ in several aspects such as the method itself, the number of tasks method considers (multi-task vs. single-task) and the subset of data considered (sub-sampling). We show that all these different aspects contribute to the success of the final ensemble. In addition, we attempt to use the drug screen data together with two novel signatures produced from the drug-induced gene expression profiles of cancer cell lines. Finally, we evaluate the method predictions by in vitro experiments in addition to the tests on data sets.The predictions of the methods, the signatures and the software are available from http://mtan.etu.edu.tr/drug-response-prediction/.

The combination of high-dimensionality and disparity of time scales encountered in many problems in computational physics has motivated the development of coarse-grained (CG) models. In this paper, we advocate the paradigm of data-driven discovery for extract- ing governing equations by employing fine-scale simulation data. In particular, we cast the coarse-graining process under a probabilistic state-space model where the transition law dic- tates the evolution of the CG state variables and the emission law the coarse-to-fine map. The directed probabilistic graphical model implied, suggests that given values for the fine- grained (FG) variables, probabilistic inference tools must be employed to identify the cor- responding values for the CG states and to that end, we employ Stochastic Variational In- ference. We advocate a sparse Bayesian learning perspective which avoids overfitting and reveals the most salient features in the CG evolution law. The formulation adopted enables the quantification of a crucial, and often neglected, component in the CG process, i.e. the pre- dictive uncertainty due to information loss. Furthermore, it is capable of reconstructing the evolution of the full, fine-scale system. We demonstrate the efficacy of the proposed frame- work in high-dimensional systems of random walkers.

People in China are the world's most optimistic when it comes to the impact of artificial intelligence on the jobs market and improving their lives, a global survey has found. Some 65 per cent of Chinese respondents believed AI and robotics would create more jobs – rather than steal them – over the next five to 10 years, according to the Digital Society Index released by UK digital marketing firm Dentsu Aegis Network on Wednesday. That compared with the global average of 29 per cent. The company polled 20,000 people across 10 countries – Australia, China, France, Germany, Italy, Japan, Russia, Spain, the United Kingdom and United States – last summer. How China's AI technology can help Twitter's suicidal users Some 71 per cent of Chinese respondents also believed that emerging digital technologies would help to solve the world's most pressing challenges such as poverty, health and environmental issues.

In the 18th century, those operating at the highest levels of society, from London to Moscow, needed to be able to speak French, then the language of status, the nobility, politics, intellectual life and modernisation. A hundred years later, British advances in industry, science and engineering meant that English succeeded French: a tongue with West Germanic origins replaced a romance language as the means of conducting business and diplomacy on the international stage. Today, even in some parts of China, English is still used as the global lingua franca, a leveller that enables deals to get done and the wheels of commerce and technology to spin. Around a decade ago, another type of language – one that was written rather than spoken – was held up as a deterministic factor for those seeking to gain influence or advantage in the digital age: coding. Its champions proselytised that proficiency in programming would determine employability and access to a thrusting, energetic entrepreneurial future.

Last night I had the opportunity to attend a debate on the future of AI at ETH university here in Zurich, Switzerland. It was a fascinating discussion between Robin Hanson, author of The Age of Em and Max Daniel of the Foundational Research Institute but I think they, like most technologists missed an important variable, humanity. The biggest difference between the two guests centered on the likelihood of Artificial General Intelligence and ability to control AI in the future -- Max arguing for thoughtful control and Robin having little fear of a Skynet scenario. This is a point that has been discussed extensively and yet in my opinion warrants additional discussion. What happens in a winner take all scenario? We move fast and break things.

We hear a lot about the dangers of video games but what if coaches used them to improve and inspire young players? Football is a simple game and nowhere more so than at youth level, where children instinctively connect the dots and know to put the round thing into the net – or between the jumpers. In coaching, we tend to worry too much about how successful teams of yesteryear were formed rather than looking forward and taking advantage of modern methods and tools. Youngsters don't live in the same world Kenny Dalglish or Denis Law grew up in. City streets are no longer littered with footballs – at least in Scotland –but while children are now restricted in ways their fathers and grandfathers weren't offline, they have greater freedom of expression and exploration online.

The thought leadership conference and exhibition AI Expo Global is set to arrive in London's Olympia on April 18-19 and will bring together AI leaders from key industries covering marketing, finance, government, public sector, healthcare, cybersecurity, HR and recruitment, automotive, industrial, developer, enterprise, and consumer sectors. With four dedicated AI conference tracks, over 12,000 attendees, 500 speakers, 300 exhibitors, and three co-located events, the AI Expo is one for technology enthusiasts and business leaders' calendars. Attendees can discover how AI is being implemented and monetized, and network with industry leaders, practitioners, and investors. Topics covered over the two-day conference include deep learning, machine learning, AI algorithms, data analytics, digital transformation, chatbots, virtual assistants, AI enterprise strategy, AI regulation and legislation, AI in the workplace, cybersecurity, AI for social good, and many more. Healthcare: Clinical trial participant identifier, preliminary diagnosis, automated image diagnosis, virtual nursing assistants, robot-assisted surgery, dosage error reduction, security, and connected machines across the industry.

Police should look at using artificial intelligence to help cope with the scale of information involved in investigations and avoid the kinds of mistakes that have led to a string of collapsed rape trials, a senior police chief said on Wednesday. Sara Thornton, the chair of the National Police Chiefs' Council, said the volume of data held by individuals had massively increased the number of potential lines of enquiry that officers must pursue to understand a case. In recent months, several rape prosecutions have been dropped after it emerged that police had failed to hand over evidence that undermined their cases. Since then, the Crown Prosecution Service has announced a review of all current rape cases and Nick Ephgrave, the NPCC's lead on criminal justice, has admitted that police have a "cultural problem" with disclosure. The attorney general's guidelines on disclosure say that police have a duty to pursue all reasonable lines of investigation, leading both towards and away from a conviction, Thornton said.

Earlier this month, when we set about to demystify some of the worst business jargon at the World Economic Forum in Davos, we could not have imagined it would hit so many of our readers' raw nerves. Hundreds felt compelled to get in touch with their own submissions, some unprintable, but the best of which we have "outlined" below. There was, of course, plenty of criticism of our selections, with many objecting to the singling out of "benchmarking" - a term that has been in use in many disciplines for several decades - and a passionate debate about the precise meaning of "negative feedback loops", more of which later. But perhaps the wittiest critique came from Charles Crowe, who maintains that "all these explanations lack granularity and do not contain metrics sufficient to let us know if we need a new paradigm". We have taken that on board, Charles.

State-space models (SSMs) are a highly expressive model class for learning patterns in time series data and for system identification. Deterministic versions of SSMs (e.g. LSTMs) proved extremely successful in modeling complex time series data. Fully probabilistic SSMs, however, are often found hard to train, even for smaller problems. To overcome this limitation, we propose a novel model formulation and a scalable training algorithm based on doubly stochastic variational inference and Gaussian processes. In contrast to existing work, the proposed variational approximation allows one to fully capture the latent state temporal correlations. These correlations are the key to robust training. The effectiveness of the proposed PR-SSM is evaluated on a set of real-world benchmark datasets in comparison to state-of-the-art probabilistic model learning methods. Scalability and robustness are demonstrated on a high dimensional problem.