BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.is.ed.ac.uk//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Scottish Centre for Diaspora Studies events
SUMMARY:Automated Shape Optimization in FEniCS - Simon Wolfgang Funke (Simula research laboratory)
DTSTART;TZID=Europe/London:20190206T160000
DTEND;TZID=Europe/London:20190206T170000
UID:TALK1714
URL:http://talks.is.ed.ac.uk/talk/1714/show
DESCRIPTION:Shape optimization constrained by partial differential equations(PDEs) are ever-present in the area of scientific computing. The PDEs considered in such applications spans a broad spectrum, ranging from steady state, linear equations, to transient non-linear equations. Efficient algorithms for shape optimization rely on shape derivatives and adjoint PDEs, which, whose derivation and implementation are known to be cumbersome and error-prone. In this talk, we present an approach where this derivation is automatically computed through high-level algorithmic differentiation tool. The shape sensitivities are computed by discrete derivatives of the mesh node sensitivities., for example structural mechanics, computational fluid dynamics and acoustics. A key advantage of our approach is that only the forward problem has to be postulated as input. Then the algorithm creates a computational tape that tracks the propagation of variables and computes corresponding gradients and Hessians. By using the operator overloading approach we inherit the parallelism and performance of the software used to solve the PDE. Our software is overloading the framework FEniCS, which has a high-level Python interface and generates parallel, optimized C code. We illustrate the efficiency and robustness of the code by presenting results for several PDEs with a wide range of solution methods.
LOCATION:JCMB5323
CONTACT:Kostas Zygalakis (kzygalak)
END:VEVENT
BEGIN:VEVENT
CATEGORIES:Scottish Centre for Diaspora Studies events
SUMMARY:From molecular mechanics to Brownian dynamics and multi-resolution methods - Radek Erban (University of Oxford)
DTSTART;TZID=Europe/London:20190313T160000
DTEND;TZID=Europe/London:20190313T170000
UID:TALK1740
URL:http://talks.is.ed.ac.uk/talk/1740/show
DESCRIPTION:Molecular dynamics (MD) approaches, based on the rules of classical mechanics, are commonly used to study the behaviour of complex biomolecules in biological applications. They are given as systems of ordinary or stochastic differential equations for the time evolution of positions and velocities of particles, representing either individual atoms or groups of atoms, describing parts of a biomolecule. One of the main limitations of all-atom MD simulations is that their direct application to the modelling of intracellular behaviour is restricted to modelling processes in relatively small domains over relatively short time intervals. In particular, intracellular processes which include transport of molecules between different parts of a cell, are usually only modelled by a much coarser modelling approach, including Brownian dynamics (BD) and other stochastic reaction-diffusion models.\nIn my talk, I will discuss connections between MD and BD, focusing on limiting theorems guaranteeing convergence of an MD model to a coarser stochastic description, given by the (generalized) Langevin dynamics written in terms of a relatively low-dimensional stochastic dynamical system. I will show how this limiting process can help us to develop and analyze multi-resolution methods for spatio-temporal modelling of intracellular processes. These methods use detailed MD simulations in localized regions of particular interest (in which accuracy and microscopic details are important) and a coarser (less-detailed) model in other regions where accuracy may be traded for simulation efficiency. Three types of multi-resolution methodologies will be considered in detail:\n(a) describing the whole biomolecule (biological structure) of interest by the detailed modelling approach which is coupled with a coarse model for the solvent molecules which are far away from the biomolecule;\n(b) describing different parts of a biomolecule by using models with different level of resolution; and\n(c) considering the region with the most detailed model as a fixed part of the physical space and allowing the biomolecule of interest to pass between this region and its surroundings, where a coarse-grained modelling approach is used.\nreferences:\n[1] R. Erban. From molecular dynamics to Brownian dynamics. Proceedings of the Royal Society A 470: 20140036 (2014)\n[2] R. Erban. Coupling all-atom molecular dynamics simulations of ions in water with Brownian dynamics. Proceedings of the Royal Society A 472: 20150556 (2016)\n[3] R. Gunaratne, D. Wilson, M. Flegg and R. Erban. Multi-resolution dimer models in heat baths with short-range and long-range interactions, to appear in Interface Focus (2019)
LOCATION:JCMB5323
CONTACT:Kostas Zygalakis (kzygalak)
END:VEVENT
BEGIN:VEVENT
CATEGORIES:Scottish Centre for Diaspora Studies events
SUMMARY:Multimodal biomedical image analysis: sparse, shallow and deep - Sotirios Tsaftaris (University of Edinburgh)
DTSTART;TZID=Europe/London:20181128T160000
DTEND;TZID=Europe/London:20181128T170000
UID:TALK1656
URL:http://talks.is.ed.ac.uk/talk/1656/show
DESCRIPTION:Medical imaging data are typically accompanied by additional information (e.g. the clinical history of the patient). At the same time, for example, magnetic resonance imaging exams typically contain more than one image modality: they show the same anatomy under different acquisition strategies revealing various pathophysiological information. The detection of disease, segmentation of anatomy and other classical analysis tasks, can benefit from a multimodal view to analysis that leverages shared information across the sources yet preserves unique (critical for diagnosis) information. It is without surprise that radiologists analyse data in this fashion, reviewing the exam as a whole. Yet, when aiming to automate analysis tasks, we still treat different image modalities in isolation and tend to ignore additional (non-image) information. In this talk, I will view modality as information extracted from the same imaging data (multimodal in the source) or from different imaging exams. I will discuss how different architectural choices can solve key problems in learning from structural information as means to reduce need for annotation. I will also show how deep neural networks can learn latent embeddings suitable for multimodal processing. I will conclude by highlighting challenges in improving our understanding and optimization of neural networks that by working across disciplines we can tackle.
LOCATION:JCMB5323
CONTACT:Kostas Zygalakis (kzygalak)
END:VEVENT
BEGIN:VEVENT
CATEGORIES:Scottish Centre for Diaspora Studies events
SUMMARY:Parameter estimation for pedestrian dynamics models - Susana Gomes (University of Warwick)
DTSTART;TZID=Europe/London:20181121T160000
DTEND;TZID=Europe/London:20181121T170000
UID:TALK1645
URL:http://talks.is.ed.ac.uk/talk/1645/show
DESCRIPTION:In this talk we present a framework for estimating parameters in macroscopic models for crowd dynamics using data from individual trajectories. We consider a model for the unidirectional flow of pedestrians in a corridor which consists of a coupling between a density dependent stochastic differential equation and a nonlinear partial differential equation for the density. In the stochastic differential equation for the trajectories, the velocity of a pedestrian decreases with the density according to the fundamental diagram. Although there is a general agreement on the basic shape of this dependence, its parametrization depends strongly on the measurement and averaging techniques used as well as the experimental setup considered. We will discuss identifiability of the parameters appearing in the fundamental diagram, introduce optimisation and Bayesian methods to perform the identification, and analyse the performance of the proposed methodology in various realistic situations. Finally, we discuss possible generalisations, including the effect of the form of the fundamental diagram and the use of experimental data.\n \n
LOCATION:JCMB5323
CONTACT:Kostas Zygalakis (kzygalak)
END:VEVENT
END:VCALENDAR