An increasing prominence of unbalanced and noisy data highlights the importance of elliptical mixture models (EMMs), which exhibit enhanced robustness, flexibility and stability over the widely applied Gaussian mixture model (GMM). However, existing studies of the EMM are typically of \textit{ad hoc} nature, without a universal analysis framework or existence and uniqueness considerations. To this end, we propose a general framework for estimating the EMM, which makes use of the Riemannian manifold optimisation to convert the original constrained optimisation paradigms into an un-constrained one. We first revisit the statistics of elliptical distributions, to give a rationale for the use of Riemannian metrics as well as the reformulation of the problem in the Riemannian space. We then derive the EMM learning framework, based on Riemannian gradient descent, which ensures the same optimum as the original problem but accelerates the convergence speed. We also unify the treatment of the existing elliptical distributions to build a universal EMM, providing a simple and intuitive way to deal with the non-convex nature of this optimisation problem. Numerical results demonstrate the ability of the proposed framework to accommodate EMMs with different properties of individual functions, and also verify the robustness and flexibility of the proposed framework over the standard GMM.

Reinforcement learning (RL) is a systematic approach to learning in sequential decision problems, where a learners' future task performance depends on its past actions. In such settings, learners have to explore, meaning they have to take actions with uncertain outcomes, to facilitate learning about the consequences of such actions. The question of how to best explore is a key open question in RL. Here, we specifically address this question from an empirical perspective, and investigate how to explore in a way that leads to sample efficient learning in deep RL, i.e., reinforcement learning with value function approximators that are parameterized as powerful neural networks. We present a surprising finding: in this setting, good approximate value functions can be learned without any explicit exploration. In fact, we find that in several cases learning without explicit exploration is equally or more sample efficient than the most-commonly used ɛ-greedy exploration scheme on several standard benchmark tasks. We present additional results that suggest a likely role of model structure (network depth and nonlinearity) in inducing such implicit exploration. We believe that our insights have strong practical implications and open up a novel line of research towards understanding exploration in deep RL.

The adoption of machine learning in high-stakes applicatio ns such as healthcare and law has lagged in part because predictions are not accomp anied by explanations comprehensible to the domain user, who often holds ult imate responsibility for decisions and outcomes. In this paper, we propose an appr oach to generate such explanations in which training data is augmented to inc lude, in addition to features and labels, explanations elicited from domain use rs. A joint model is then learned to produce both labels and explanations from the inp ut features. This simple idea ensures that explanations are tailored to the compl exity expectations and domain knowledge of the consumer. Evaluation spans multipl e modeling techniques on a simple game dataset, an image dataset, and a chemi cal odor dataset, showing that our approach is generalizable across domains a nd algorithms. Results demonstrate that meaningful explanations can be reli ably taught to machine learning algorithms, and in some cases, improve modeling ac curacy.

I completed my PhD in astrophysics at the University of Sheffield, UK, focusing on studying galaxy formation and evolution. After my PhD I did a number of post doctoral position until I jumped to the private sector early in 2014. Currently, I head the data science team at Simply Business, and insurance company. Previously I have worked in the fashion and retail industries developing recommendation algorithms. In addition I am also an advisor at jaggu.com, a start-up focused on data enrichment and recommendation systems in the retail space.

When will machines have human agility? That's what the film studio Universal Everything tries to answer in their captivating videos pairing a dancer and a copycat robot mimicking his moves. Set in a spacious, well-worn dance studio, a dancer teaches a series of robots how to move. As the robots' abilities develop from shaky mimicry to composed mastery, a physical dialogue emerges between man and machine – mimicking, balancing, challenging, competing, outmanoeuvring. Can the robot keep up with the dancer?

An affordable British-built self-driving family car that can steer and brake by itself has been launched in the UK today by Japanese car giant Nissan. It has fitted its school-run favourite Nissan Qashqai sports utility vehicle with advanced autonomous driving technology that until now has been the preserve of expensive luxury vehicles such as top of the range BMW and Mercedes-Benz models. We were among the first to try the latest model - and the autonomous driving modes - on and off-road before it hits showrooms. Britain's first semi-autonomous car: The Sunderland-built Qashqai SUV will be the first UK-assembled model designed for families that will feature a host of new driverless tech Equipped with Nissan's new'ProPilot' system, the smart Qashqai, built at its UK factory in Sunderland, can steer, accelerate and brake automatically using data supplied by a camera mounted on its windscreen and a radar behind badge on front grille. Nissan stresses that its self-driving technology is currently there to aid the driver, not to take over from him or her. It added that the system should help transform driving by making it'less stressful and more confident'.

Upon closer inspection, the satellite had noticed that an area that should have been shrouded in forest, was now barren. Within hours, a call had been made to a global conservation group, who mounted a legal case against the logging companies operating in the area. That process, historically, could have taken months of observing and recording changes. What's more, in remote areas such as the Ussuri Taiga in Russia's Far East, policing illegal logging operations have historically had little impact on the extraction of timber. But thanks to artificial intelligence (AI) and satellites, the ability to observe and respond to changes has become much faster.

Gini is dedicated to putting an end to paperwork. Voted as FinTech Startup of the Year 2015, it uses artificial intelligence to capture, structure and extract data from documents for further usage. Based on that, Gini provides smart solutions on how to get paperwork done with minimal effort, allowing customers to use their time for things that really matter in life. Gini's photo payment is already a standard in the German online banking market, used by millions of banking customers.

AI-based decisions and tools are starting to have a profound impact on people's lives and insurers' businesses. To learn, AI must consume a lot of data. But what if the data fed to an AI solution was biased? Can human-machine collaboration mitigate the potential risks? Citizen AI: Raising AI to Benefit Business and Society is one of five trends highlighted in Accenture's Technology Vision for Insurance 2018.

'Artificial intelligence software capable of interpreting images, matching faces and analysing patterns of communication is being piloted by UK police forces to speed up examination of mobile phones seized in crime investigations. Cellebrite, the Israeli-founded and now Japanese-owned company behind some of the software, claims a wider rollout would solve problems over failures to disclose crucial digital evidence that have led to the collapse of a series of rape trials and other prosecutions in the past year. However, the move by police has prompted concerns over privacy and the potential for software to introduce bias into processing of criminal evidence. As police and lawyers struggle to cope with the exponential rise in data volumes generated by phones and laptops in even routine crime cases, the hunt is on for a technological solution to handle increasingly unmanageable workloads. Some forces are understood to have backlogs of up to six months for examining downloaded mobile phone contents.