19 February 2020
Professor Chun-Hung Chen (George Mason), hosted by Juergen Branke
Efficient Simulation-based Optimization with Optimal Computing Budget Allocation and Smart Space Transformation
Simulation is a powerful modeling and software tool for analyzing modern complex systems that arise in manufacturing, power grids, transportation, healthcare, finance, defense, and many other fields. Detailed dynamics of complex, stochastic systems can be modeled in simulation. This capability complements the inherent limitation of traditional optimization, so the combining use of simulation and optimization is growing in popularity. This seminar discusses the computational issues in such a combination, and presents our effective approaches. Our approach aims to maximize the efficiency of finding a good solution, via optimally searching the solution space and sampling the simulation replications. Further, we propose to transform the solution space into a smart space which is smoother and has nice properties. Thus, the search becomes easier and more efficient in the transformed space. A key component of our methodologies is a new technique called Optimal Computing Budget Allocation (OCBA) initially developed by the speaker.
12 February 2020
Dr Raluca Eftimie (Dundee), hosted by David Rand
Modelling viral therapies for cancer: anti-viral vs anti-tumour immunity
Over the past years oncolytic viruses have generated much interest in cancer therapy, mainly due to the fact that once a virus is injected into the patient it can actively search for cancer cells and destroy them. However, the anti-tumour effect of oncolytic viruses is greatly diminished by anti-viral immune responses, as well as by physical barriers inside the tumour.
Using various single-scale and multi-scale modelling approaches, we will investigate the delicate balance between anti-viral and anti-tumour immune responses in the context of virus-tumour-immune interactions. We will also discuss the effect of extracellular matrix on the spread of oncolytic viruses.
05 February 2020
Dr Xavier Didelot (Warwick) hosted by Mike Tildesley
Modelling recombination in bacterial genomic evolution
Recombination happens frequently in most bacterial species. Traditional phylogenetic techniques do not account for this, which can greatly limit their usefulness for the analysis of genomic data. The coalescent with gene conversion accurately models the ancestry process of bacteria, and this can be used to simulate realistic data, but it is too complex to use in an inferential setting. Approximations have therefore been introduced, which are centred around the concept of the clonal genealogy, that is the phylogeny obtained by following the line of ancestry of the recipient of each recombination event. I will review these mathematical models and ongoing efforts to develop statistical methods and software to perform phylodynamic analysis in recombining bacteria.
29 January 2020
Dr Vishwesh Kulkarni (Warwick) hosted by Colm Connaughton
22 January 2020
Dr Simon Spencer (Warwick) hosted by Magnus Richarson
Bayesian methods for spatio-temporal modelling of campylobacteriosis
Campylobacteriosis is a common form of food poisoning with a complex epidemiology. Despite the large number of cases, the dominant pathways to infection were until recently poorly understood. In this talk I will outline how novel statistical methodology has helped to develop our understanding of this pathogen as part of a multidisciplinary approach spanning epidemiology, genetics, public health and evolutionary biology. In particular I will outline how to identify risk factors from the spatial distribution of cases and how to detect outbreaks from a background of sporadic cases.
Joint work with Nigel French.
20th November 2019
Dr Jean-Baptiste Mouret (University de Lorraine) hosted by Juergen Branke
Gaussian processes for fast adaptation in robotics
Modern robots are more and more robust, but they are still fragile machines that often stop functioning when an unforeseen situation arises, for instance when they break a mechanical part. One way to give recovery abilities to robots is to allow them to learn by trial-and-error a compensatory behaviour; however, this requires learning algorithms that can find solutions in less than a dozen of trials.
13 November 2019
Dr Chandrasekhar Venkataraman (University of Sussex) hosted by Bjorn Stinner
Modelling a cereal killer: A pressure sensing kinase aids rice plant infection by the blast fungus Magnaporthe oryzae.
Abstract: The fungus Magnaporthe oryzae generates enormous turgor pressures allowing it to puncture rice leaves and subsequently to infect the entire plant. The fungus has devastating consequences on the global rice yield with rice blast disease accounting for approximately 30% of rice production losses globally. In this talk we develop a mathematical model for the puncturing of the rice leaf by the fungus. The model couples an evolution law for the growth of a specialised cell known as an appressorium to a reaction diffusion system that holds on the surface of the appressorium We derive a finite element method to approximate solutions to the model and we show some computational results. Using the model in combination with experimental observations we discover the central role played by a pressure sensitive kinase in the infection process.
30 October 2019
Javier Gonzales (Amazon) hosted by Juergen Branke
Gaussian process and the common ground of decision making under uncertainty
Abstract: Rather than focusing on one single method or problem, in this talk we will review the common ground of several decision making methods under uncertainty (such as Bandits, Bayesian optimization, Active Learning and Bayesian quadrature) and the role that Gaussian processes play as belief model in these approaches. We will discuss some recent advances in these fields and propose a general recipe to design and implement new methods.
Dr Richard Everitt (University of Warwick) hosted by Stefan Grosskinsky
ABC for expensive simulators
Abstract: Approximate Bayesian computation (ABC) is now an established technique for statistical inference in the form of a simulator, and approximates the likelihood at a parameter θ by simulating auxiliary data sets x and evaluating the distance of x from the true data y. Synthetic likelihood (SL) is a related approach that uses simulated auxiliary data sets to contract a Gaussian approximation to the likelihood. However, these approaches are not computationally feasible in cases where using the simulator for each θ is very expensive. This talk introduces the main concepts in ABC and SL, and describes strategies for inference when the simulator is expensive. Applications to stochastic differential equation models and individual based models will be presented.
References: delayed acceptance ABC-SMC (https://arxiv.org/abs/1708.02230); bootstrapped synthetic likelihood (https://arxiv.org/abs/1711.05825); rare event ABC (https://arxiv.org/abs/1611.02492).