Past Industrial and Interdisciplinary Workshops

2 December 2016
10:00
Christian Sommeregger & Wen Wong
Abstract

Hotels.com is one of the world’s leading accommodation booking websites featuring an inventory of around 300.000 hotels and 100s of millions of users. A crucial part of our business is to act as an agent between these two sides of the market, thus reducing search costs and information asymmetries to enable our visitors to find the right hotel in the most efficient way.

From this point of view selling hotels is one large recommendation challenge: given a set of items and a set of observed choices/ratings, identify a user’s preference profile.  Over the last years this particular problem has been intensively studied by a strongly interdisciplinary field based on ideas from choice theory, linear algebra, statistics, computer science and machine learning. This pluralism is reflected in the broad array of techniques that are used in today’s industry applications, i.e. collaborative filtering, matrix factorization, graph-based algorithms, decision trees or generalized linear models.

The aim of this workshop is twofold.

Firstly we want to give some insight into the statistical modelling techniques and assumptions employed at hotels.com, the practical challenges one has to face when designing a flexible and scalable recommender system and potential gaps between current research and real-world applications.

Secondly we are going to consider some more advanced questions around learning to rank from partial/incomplete feedback (1), dealing with selection-bias correction (2) and how econometrics and behavioral theory (eg Luce, Kahneman /Tversky) can be used to complement existing techniques (3).

 

  • Industrial and Interdisciplinary Workshops
25 November 2016
10:00
Paul Sweeney
Abstract

Environmental risk assessments for chemicals in the EU rely heavily upon modelled estimates of potential concentrations in soil and water.  A key parameter used by these models is the degradation of the chemical in soil which is derived from a kinetic fitting of laboratory data using standard fitting routines.  Several different types of kinetic can be represented such as: Simple First Order (SFO), Double First Order in Parallel (DFOP), and First Order Multi-Compartment (FOMC). Choice of a particular kinetic and selection of a representative degradation rate can have a huge influence on the outcome of the risk assessment. This selection is made from laboratory data that are subject to experimental error.  It is known that the combination of small errors in time and concentration can in certain cases have an impact upon the goodness of fit and kinetic predicted by fitting software.  Syngenta currently spends in the region of 4m GBP per annum on laboratory studies to support registration of chemicals in the EU and the outcome of the kinetic assessment can adversely affect the potential registerability of chemicals having sales of several million pounds.  We would therefore like to understand the sensitivities involved with kinetic fitting of laboratory studies.  The aim is to provide guidelines for the conduct and fitting of laboratory data so that the correct kinetic and degradation rate of chemicals in environmental risk assessments is used.

  • Industrial and Interdisciplinary Workshops
11 November 2016
10:00
Harry McEvoy
Abstract

Dstl are interested in removing liquid contaminants from capillary features (cracks in surfaces, screw threads etc.). We speculated that liquid decontaminants with low surface tension would have beneficial properties. The colloid literature, and in particular the oil recovery literature, discusss the properties of multiphase systems in terms of “Winsor types”, typically consisting of “brine” (water + electrolyte), “oil” (non-polar, water-insoluble solvent) and surfactant. Winsor I systems are oil-in-water microemulsions and Winsor II systems are water-in-oil microemulsions. Under certain circumstances, the mixture will separate into three phases. The middle (Winsor III) phase is surfactant-rich, and is reported to exhibit ultra-low surface tension. The glycol ethers (“Cellosolve” type solvents) consist of short (3-4) linked ether groups attached to short (3-4 carbon) alkyl chains. Although these materials would not normally be considered to be surfactants, their polar head, non-polar tail properties allow them to form a “surfactantless” Winsor III middle phase. We have found that small changes in temperature, electrolyte concentration or addition of contaminant can cause these novel colloids to phase separate. In our decontamination experiments, we have observed that contaminant-induced phase separation takes the form of droplets of the separating phase. These droplets are highly mobile, exhibiting behaviour that is visually similar to Brownian motion, which induces somewhat turbulent liquid currents in the vicinity of the contaminant. We tentatively attribute this behaviour to the Marangoni effect. We present our work as an interesting physics/ physical chemistry phenomenon that should be suitable for mathematical analysis.

  • Industrial and Interdisciplinary Workshops
4 November 2016
10:00
Gil travish
Abstract

Currently all medical x-ray imaging is performed using point-like sources which produce cone or fan beams. In planar radiology the source is fixed relative to the patient and detector array and therefore only 2D images can be produced. In CT imaging, the source and detector are rotated about the patient and through reconstruction (such as Radon methods), a 3D image can be formed. In Tomosynthesis, a limited range of angles are captured which greatly reduces the complexity and cost of the device and the dose exposure to the patient while largely preserving the clinical utility of the 3D images. Conventional tomosynthesis relies on mechanically moving a source about a fixed trajectory (e.g. an arc) and capturing multiple images along that path. Adaptix is developing a fixed source with an electronically addressable array that allows for a motion-free tomosynthesis system. The Adaptix approach has many advantages including reduced cost, portability, angular information acquired in 2D, and the ability to shape the radiation field (by selectively activating only certain emitters).


The proposed work would examine the effects of patient motion and apply suitable corrections to the image reconstruction (or raw data). Many approaches have been considered in the literature for motion correction, and only some of these may be of use in tomosynthesis. The study will consider which approaches are optimal, and apply them to the present geometry.


A related but perhaps distinct area of investigation is the use of “structured light” techniques to encode the x-rays and extract additional information from the imaging. Most conventional structured light approaches are not suitable for transmissive operation nor for the limited control available in x-rays. Selection of appropriate techniques and algorithms, however, could prove very powerful and yield new ways of performing medical imaging.


Adaptix is a start-up based at the Begbroke Centre for Innovation and Enterprise. Adaptix is transforming planar X-ray – the diagnostic imaging modality most widely used in healthcare worldwide. We are adding low-dose 3D capability – digital tomosynthesis - to planar X-ray while making it more affordable and truly portable so radiology can more easily travel to the patient. This transformation will enhance patient’s access to the world’s most important imaging technologies and likely increases the diagnostic accuracy for many high incidence conditions such as cardiovascular and pulmonary diseases, lung cancer and osteoporosis. 
 

  • Industrial and Interdisciplinary Workshops
17 June 2016
10:00
Abstract

T cells are important white blood cells that continually circulate in the body in search of the molecular signatures ('antigens') of infection and cancer. We (and many other labs) are trying to construct models of the T cell signalling network that can be used to predict how ligand binding (at the surface of the cell) controls gene express (in the nucleus). To do this, we stimulate T cells with various ligands (input) and measure products of gene expression (output) and then try to determine which model must be invoked to explain the data. The challenge that we face is finding 1) unique models and 2) scaling the method to many different input and outputs.

  • Industrial and Interdisciplinary Workshops
10 June 2016
10:00
Alexander Denev
Abstract

Markit is a leading global provider of financial information services. We provide products that enhance transparency, reduce risk and improve operational efficiency.

We wish to find ways to automatically detect and label ‘extreme’ occurrences in a time series such as structural breaks, nonlinearities, and spikes (i.e. outliers). We hope to detect these occurrences in the levels, returns and volatility of a time series or any other transformation of it (e.g. moving average).

We also want to look for the same types of occurrences in the multivariate case in a set of time series through measures such as e.g. correlations, eigenvalues of the covariance matrix etc. The number of time series involved is of the order 3x10^6.

We wish to explain the appearance of an ‘extreme’ occurrence or a cluster of occurrences endogenously, as an event conditional on the values of the time series in the set, both contemporaneously and/or as conditional on their time lags.

Furthermore, we would like to classify the events that caused the occurrence in some major categories, if found e.g. shock to oil supply, general risk aversion, migrations etc. both algorithmically and by allowing human corrective judgement (which could become the basis for supervised learning).

  • Industrial and Interdisciplinary Workshops
3 June 2016
10:00
Abstract

We are entering a world where unmanned vehicles will be common. They have the potential to dramatically decrease the cost of services whilst simultaneously increasing the safety record of whole industries.

Autonomous technologies will, by their very nature, shift decision making responsibility from individual humans to technology systems. The 2010 Flash Crash showed how such systems can create rare (but not inconceivably rare) and highly destructive positive feedback loops which can severely disrupt a sector.

In the case of Unmanned Air Systems (UAS), how might similar effects obstruct the development of the Commercial UAS industry? Is it conceivable that, like the high frequency trading industry at the heart of the Flash Crash, the algorithms we provide UAS to enable autonomy could decrease the risk of small incidents whilst increasing the risk of severe accidents? And if so, what is the relationship between probability and consequence of incidents?

  • Industrial and Interdisciplinary Workshops
27 May 2016
10:00
Abstract

We aim to determine how cells faithfully complete genome replication. Accurate and complete genome replication is essential for all life. A single DNA replication error in a single cell division can give rise to a genomic disorder. However, almost all experimental data are ensemble; collected from millions of cells. We used a combination of high-resolution, genomic-wide DNA replication data, mathematical modelling and single cell experiments to demonstrate that ensemble data mask the significant heterogeneity present within a cell population; see [1-4]. Therefore, the pattern of replication origin usage and dynamics of genome replication in individual cells remains largely unknown. We are now developing cutting-edge single molecule methods and allied mathematical models to determine the dynamics of genome replication at the DNA sequence level in normal and perturbed human cells.

[1] de Moura et al., 2010, Nucleic Acids Research, 38: 5623-5633

[2] Retkute et al, 2011, PRL, 107:068103

[3] Retkute et al, 2012, PRE, 86:031916

[4] Hawkins et al., 2013, Cell Reports, 5:1132-41

  • Industrial and Interdisciplinary Workshops

Pages