Dr Man Lai Tang joined Brunel in 2021; his research focuses on surveys with sensitive questions, quantile regression modelling and profile likelihood methods. His talk will be on Statistics and surveys with sensitive questions.

ABSTRACT: Although the item count technique is very useful in surveys with sensitive questions, privacy of those respondents who possess the sensitive characteristic of interest may not be well protected due to a defect in its original design. In this talk, a new survey designs called the Poisson item count technique, which replaces several independent Bernoulli random variables required by the original item count technique with a single Poisson random variable, will be introduced. The proposed models not only provide closed form variance estimate and confidence interval within [0, 1] for the sensitive proportion, but also simplify the survey design of the original item count technique. Most importantly, respondents’ privacy will be well protected by the new design. Some important extensions and future work will be discussed.

Dr Sébastien Guenneau is a Director of Research at the Centre National de la Recherche Scientifique (CNRS) working in the CNRS-Imperial Abraham de Moivre Unité Mixte Internationale. His research interests lie in the physics of metamaterials for an enhanced control of wave and diffusion phenomena (including negative refraction and cloaking), in homogenization of multiscale periodic and quasi-periodic media, and in finite element models of acoustic, electromagnetic, hydrodynamic and elastodynamic waves. His talk will be on Invisibility cloaks across the scales.

ABSTRACT: In this talk, I will review some research advances in photonics in the past twenty years that might lead to paradigm shift in earthquake engineering in the near future. The physicist Sir John Pendry proposed a concept of a convergent flat lens with a negative refractive index allowing for space folding at the turn of the millennium. Six years later, he came up with the breakthrough concept of invisibility cloaks designed via space transformations. Pendry’s lens and cloak have been fabricated and experimentally tested using man-made metamaterials. A metamaterial (from the Greek word meta, meaning beyond and the Latin word materia, meaning material) is any material engineered on a subwavelength scale to have an effective property that is not found in naturally occurring materials, such as a negative index of refraction!

Surprisingly, metamaterials can be scaled up, their funny effective properties being mostly scale invariant! This makes it possible to bridge the fields of photonics and geophysics: for instance the London Millennium bridge may have an effective negative mass upon resonance induced by pedestrian walk, which is why it was temporarily closed just after the opening ceremony until this matter was solved (using passive mass dampers placed underneath the bridge). Ten years ago, two large scale experiments on seismic metamaterials demonstrated a cloaking effect for surface Rayleigh waves generated by a time-harmonic source at 50 Hertz in a sedimentary soil structured with boreholes 0.3m in diameter and a lensing effect via negative refraction at 10 Hertz for surface Rayleigh waves generated by a multi-frequency source in a sedimentary soil structured with boreholes 2m in diameter. These experiments have fuelled the interest in large scale mechanical metamaterials for applications in civil and maritime engineering.

My talk will thus cover a brief history of time (20 years of research in metamaterials is short compared to over 500 years of research in Newtonian mechanics!) with just a few mathematical equations, but more importantly many applications in amazing physical effects (such as making a building invisible to ocean waves and earthquakes)!

Dr Martins Bruveris is a Senior Applied Scientist at Onfido, which he joined from Brunel, where he was previously a lecturer in the Mathematics Department. Martins is a machine learning researcher, with a background in mathematics and differential geometry. His talk will discuss how to train a face recognition system.

ABSTRACT: Face recognition systems are becoming more and more ubiquitous. This talk will show you how they work under the hood: how the availability of large face datasets combined with deep learning allows one to train face recognition systems that far surpass human accuracy; and how automated face recognition raises questions of racial bias and what can be done to mitigate it.

Launch Event: Centre for Mathematical and Statistical Modelling