Seminars this semester


   Series:


May 13 Mon Natasha Ellison, Sara Hilditch, Elena Marensi, Alison Parton, Lizzie Sheppeck, Sarah Whitehouse, Sadiah Zahoor (University of Sheffield) International Women in Mathematics Day
10:00 LT-5
Click here to insert in your Google calendar

May 13 Mon Suzana de Souza e Almeida Silva (Technological Institute of Aeronautics, Sao Paulo) Plasma Dynamics Group
13:00 LT9 (Hicks Building) Lagrangian Coherent Structures: Overview and applications in solar physics
 
  Abstract:
Lagrangian coherent structures (LCS) is a newly developed theory which describes the skeleton of turbulent flows. LCS act as barriers in the flow, separating regions with different dynamics and organizing the flow into coherent patterns. This talk will introduce some concepts of LCT techniques as well as recent application to solar physics problems.
Click here to insert in your Google calendar

May 13 Mon Nuria Folguera Blasco (Crick Institute) Mathematical Biology Seminar Series
14:00 Hicks F20
Click here to insert in your Google calendar

May 14 Tue Anna Krystalli / Alison Parton / Lyn Taylor (Sheffield / Sheffield / Phastar) RSS Seminar Series
16:00 LT 5 Putting the R in Reproducible Research / Cloud Computing with R / R Validation Hub Project
 
  Abstract:
R and its ecosystem of packages offers a wide variety of statistical and graphical techniques and is increasing in popularity as the tool of choice for data analysis in academia. In addition to its powerful analytical features, the R ecosystem provides a large number of tools and conventions to help support more open, robust and reproducible research. This includes tools for managing research projects, building robust analysis workflows, documenting data and code, testing code and disseminating and sharing analyses. In this talk we’ll take a whistle-stop tour of the breadth of available tools, demonstrating the ways R and the Rstudio integrated development environment can be used to underpin more open reproducible research and facilitate best practice.


R has cemented itself as the language of choice for many a statistician and data scientist, but is often heckled as a sluggish competitor to the likes of python. This talk will discuss one avenue for maintaining the comfort and power of R (see Anna’s talk!) without having to wait days for your desktop analysis to complete. The foreach package is a set of functions that allow virtually anything that can be expressed as a for-loop as a set of parallel tasks. By registering a parallel backend through the doParallel package, you can speed up the run-time of your work by utilising the full capacity of your machine. I’ll introduce how to rewrite workflows to utilise the foreach approach and show how you can implement a parallel workflow on your own machine with doParallel. For a low-range machine, this will reduce your run-time by 4-fold and for those lucky few with high-range budgets you’ll receive something around 16-fold. So how about going one step further, and increasing to hundreds-fold? We can achieve this by using cloud computing services, taking the load away from your own machine. Cloud computing services have been seen to have a steep learning curve and this has led to many shying away from using such a useful resource. I’ll introduce you to the doAzureParallel package for R, create by Microsoft to bypass this learning curve and allow you to implement the foreach package in parallel in the cloud with only minor amendments to the R code that has been blighting you for months.


To date, the use of R Software in the pharmaceutical industry has been relatively limited to exploratory work and not routinely used in regulatory submissions where SAS® Software is still favored. One of the difficulties in using R for submissions is being able to provide the regulators with appropriate documentation of testing and validation for the packages used. In June 2018 the R consortium granted funding for a PSI AIMS SIG initiative to create an online ‘R package validation repository’. With representatives from Abbvie, Amgen, Biogen, Eli Lilly, FDA, GSK, J&J, Merck, Merck KGaA, Novartis, PPD, PRA, Pfizer, Roche / Genentech, Syne qua non and the Transcelerate project, the ‘R Validation Hub’ team launched a free to access web site to host validation documentation and metrics for R packages (https://www.pharmar.org/). Although, the project is still in its early stages, we are looking to expand on the website content and encourage contribution of R metrics and tests for packages from all R-users. The talk will discuss what is meant by validation, how R differs to SAS, justify our approach to the validation issue and present the future capabilities of the website and how all R-users are set to benefit from the work.
Click here to insert in your Google calendar

May 15 Wed Martina Balagovic (Newcastle) Pure Maths Colloquium
14:00 J11 Quantum Yang Baxter equation, the reflection equation, and their universal solutions
 
  Abstract:

The quantum Yang Baxter equation arose in statistical mechanics around 1970 as the consistency condition for an interaction of two particles on a line. In the 1980s, Drinfled and Jimbo introduced quantum groups (deformations of universal enveloping algebras of Lie algebras), and showed that they allow a universal R matrix - an element constructed from the algebra, which systematically produces a solution of the quantum Yang Baxter equation in every representation of this algebra. In turn, this imposes a structure of a braided tensor category on representations of the quantum group (i.e. gives an action of the braid group of type A) and leads to the Reshetikhin-Turaev construction of invariants of knots, braids, and ribbons.

Considering the same problem with a boundary (on a half line instead of a line) leads to the consistency condition called the (quantum) reflection equation, introduced by Cherednik and Sklyanin in the 1980s. I will explain how, in the joint work with S. Kolb, we use quantum symmetric pairs (Noumi, Sugitani, and Dijkhuizen; Letzter 1990s) to construct a universal K-matrix - an element which systematically produces solutions of the reflection equation. This gives an action of the braid group of type B, endowing the corresponding category of representations with a structure of a braided tensor category with a cylinder twist (as defined by T. tom Dieck, R. Haring-Oldenburg 1990s).

Click here to insert in your Google calendar

May 16 Thu Christopher Fallaize (Nottingham) Statistics Seminar
14:00 LT E Unlabelled Shape Analysis with Applications in Bioinformatics
 
  Abstract:
In shape analysis, objects are often represented as configurations of points, known as landmarks. The case where the correspondence between landmarks on different objects is unknown is called unlabelled shape analysis. The alignment task is then to simultaneously identify the correspondence between landmarks and the transformation aligning the objects. In this talk, I will discuss the alignment of unlabelled shapes, and discuss two applications to problems in structural bioinformatics. The first is a problem in drug discovery, where the main objective is to find the shape information common to all, or subsets of, a set of active compounds. The approach taken resembles a form of clustering, which also gives estimates of the mean shapes of each cluster. The second application is the alignment of protein structures, which will also serve to illustrate how the modelling framework can incorporate very general information regarding the properties we would like alignments to have; in this case, expressed through the sequence order of the points (amino acids) of the proteins.
Click here to insert in your Google calendar

May 16 Thu Peter Wyper (University of Durham) Plasma Dynamics Group
16:00 Room K14 (Hicks Building) Reconnection, Topology and Solar Eruptions
 
  Abstract:
The majority of free energy in the solar corona is stored within sheared magnetic field structures known as filament channels. Filament channels spend most of their life in force balance before violently erupting. The largest produce powerful solar flares and coronal mass ejections (CMEs), whereby the filament channel is bodily ejected from the Sun. However, a whole range of smaller eruptions and flares also occur throughout the corona. Some are ejective, whilst others are confined. Recently it has been established that coronal jets are also typically the result of a filament channel eruption. The filament channels involved in jets are orders of magnitude smaller than the ones which produce CMEs. In this talk I will start by considering these tiny, jet producing eruptions. I will introduce our MHD simulation model that well describes them and then discuss what jets can tell us about solar eruptions in general. Specifically, I will argue that many different types of eruption can be understood by considering two defining features: the scale of the filament channel and its interaction via reconnection with its surrounding magnetic topology.
Click here to insert in your Google calendar

May 17 Fri Gong Show Topology seminar
16:00 J11
Click here to insert in your Google calendar

May 30 Thu David Jess (Queen's University (Belfast)) Plasma Dynamics Group
14:00 Room LT10 (Hicks Building) Resonance Cavities: A wave amplification mechanism above highly magnetic sunspots
 
  Abstract:
The solar atmosphere provides a unique astrophysical laboratory to study the formation, propagation, and subsequent dissipation of magnetohydrodynamic (MHD) waves across a diverse range of spatial scales. The concentrated magnetic fields synonymous with sunspots allow the examination of guided magnetoacoustic modes as they propagate upwards into the solar corona, where they exist as ubiquitous 3-minute waves readily observed along loops, plumes and fan structures. While cutting-edge observations and simulations are providing insights into the underlying wave generation and damping mechanisms, the in-situ amplification of magnetoacoustic waves as they propagate through the solar chromosphere has proved difficult to explain. Here we provide observational evidence of a resonance cavity existing above a magnetic sunspot, where the intrinsic temperature stratification provides the necessary atmospheric boundaries responsible for the resonant amplification of these waves. Through comparisons with high-resolution numerical MHD simulations, the geometry of the resonance cavity is mapped across the diameter of the underlying sunspot, with the upper boundaries of the chromosphere ranging between 1300–2300 km. This brings forth important implications for next-generation ground-based observing facilities, and provides an unprecedented insight into the MHD wave modelling requirements for laboratory and astrophysical plasmas.
Click here to insert in your Google calendar

May 31 Fri Prof Yuanyong Deng (Director of Huairoi Observatory) (NAOC, CAS, China) SP2RC seminar
13:00 LT09 The observational research and projects of solar physics in China
 
  Abstract:
In this presentation I will briefly introduce recent solar observation and related research in China. Up to now all these observations come from ground-based telescopes. In the near future, Chinese will have our first space solar observatory by the Advanced Space solar telescope (ASO-S). In addition to ASO-S, some other projects under development or proposed will also be introduced and discussed.
Click here to insert in your Google calendar

Jun 11 Tue Andreas Krug (Magburg) Algebra / Algebraic Geometry seminar
14:00 J11 Stability of Tautological Bundles on Symmetric Products of Curves
 
  Abstract:
Given a vector bundle E over smooth variety X, there is a natural way to associate a vector bundle, called tautological bundle, on the Hilbert scheme of points on X. In this talk, we will discuss stability of tautological bundles in the case that X is a curve.
Click here to insert in your Google calendar

Jun 18 Tue Tong Liu (Purdue) Number Theory seminar
14:00 J11 p-divisible groups and crystalline representations over relative base
 
  Abstract:
Let K be a p-adic field, it is known that p-adic Tate module of p-divisible group over O_K is crystalline representation with Hodge-Tate weights in [0, 1]. And conversely any such crystalline representation arise from a p-divisible group over O_K. In this talk, we discuss how to generalize this result to relative bases when O_K is replaced by more general rings, like, Z_p[[t]]. This is a joint work with Yong Suk Moon.
Click here to insert in your Google calendar

Jun 20 Thu Ioannis Kontogiannis (Leibniz-Institut für Astrophysik Potsdam, AIP)
10:00 LT11 Emergence of small-scale magnetic flux in the quiet Sun, observed from the photosphere to the corona
 
  Abstract:
We study the emergence and evolution of new magnetic flux in the vicinity of a quiet Sun network. We employ high-resolution spectropolarimetric, spectroscopic and spectral imaging observations from ground-based (Dutch Open Telescope) and space-born instruments (TRACE, Hinode, SoHO), which provided a multi-wavelength, tomographic view of the region from the photosphere up to the corona. Throughout its evolution, the region exhibited many of the phenomena revealed by recent simulations. The event starts with a series of granular-scale events, which follow the photospheric flow field and merge to form a small-scale magnetic flux system of the order of 1018 Mx. Spectropolarimetric inversions reveal an evolving, complicated pattern of horizontal and vertical magnetic field patches at the region between the main polarities. As the magnetic flux accumulates and the region expands, Doppler-shifted H-alpha absorption features appear above and at the crests of the structure, indicating an immediate interaction with the pre-existing, overlying magnetic field. Roughly 60 min after the region first emerged at the photosphere, a jet-like feature appeared in the chromosphere and a small soft X-ray bright point formed in the corona. The coronal brightening exhibited intense spatial and temporal variations and had a lifetime that exceeded one hour. EUV spectroscopy and DEM analysis revealed temperatures up to 106 K and densities up to 1010 cm-3. Even in the absence of a strong ambient magnetic field, small-scale magnetic flux emergence affects dramatically the dynamics and shape of the quiet Sun.
Click here to insert in your Google calendar

Jun 20 Thu Ioannis Kontogiannis (Leibniz-Institut für Astrophysik Potsdam, AIP) SP2RC seminar
10:00 LT11 Emergence of small-scale magnetic flux in the quiet Sun, observed from the photosphere to the corona
 
  Abstract:
We study the emergence and evolution of new magnetic flux in the vicinity of a quiet Sun network. We employ high-resolution spectropolarimetric, spectroscopic and spectral imaging observations from ground-based (Dutch Open Telescope) and space-born instruments (TRACE, Hinode, SoHO), which provided a multi-wavelength, tomographic view of the region from the photosphere up to the corona. Throughout its evolution, the region exhibited many of the phenomena revealed by recent simulations. The event starts with a series of granular-scale events, which follow the photospheric flow field and merge to form a small-scale magnetic flux system of the order of 1018 Mx. Spectropolarimetric inversions reveal an evolving, complicated pattern of horizontal and vertical magnetic field patches at the region between the main polarities. As the magnetic flux accumulates and the region expands, Doppler-shifted H-alpha absorption features appear above and at the crests of the structure, indicating an immediate interaction with the pre-existing, overlying magnetic field. Roughly 60 min after the region first emerged at the photosphere, a jet-like feature appeared in the chromosphere and a small soft X-ray bright point formed in the corona. The coronal brightening exhibited intense spatial and temporal variations and had a lifetime that exceeded one hour. EUV spectroscopy and DEM analysis revealed temperatures up to 106 K and densities up to 1010 cm-3. Even in the absence of a strong ambient magnetic field, small-scale magnetic flux emergence affects dramatically the dynamics and shape of the quiet Sun.
Click here to insert in your Google calendar

Jul 2 Tue Evgeny Shinder (Sheffield)
15:00 Hicks LT 2 Geometry of Singularities
Click here to insert in your Google calendar

Jul 2 Tue Caitlin Buck (Sheffield)
15:30 Hicks, LT2 Maintaining Impact in Interdisciplinary Research
Click here to insert in your Google calendar

Jul 2 Tue Neil Dummigan (Sheffield)
16:30 Hicks LT2 Lattices and theta series
Click here to insert in your Google calendar

Aug 14 Wed Ben Evans (Bristol) Mathematical Biology Seminar Series
16:00 Alfred Denny Conference Room Building biological constraints into convolutional neural networks for classification overcomes biases within datasets
 
  Abstract:
Part of the appeal of deep convolutional networks is their ability to learn on raw data, obviating the need to hand-code the feature space. It has been demonstrated that when networks perform "end- to-end" learning, they develop features in early layers that not only lead to a good classification performance but also resemble the representations found in biological vision systems. These results have been used to draw various parallels between deep learning systems and human visual perception. In this study, we show that end-to-end learning in standard convolutional neural networks (CNNs) trained on a modified CIFAR-10 dataset are found to rely upon idiosyncratic features within the dataset. Instead of relying on abstract features such as object shape, end-to-end learning can pick up on low-level and spatially high-frequency features, such as noise-like masks. Such features are extremely unlikely to play any role in human object recognition, where instead a strong preference for shape is observed. Through a series of empirical studies, we show that these CNNs cannot overcome such problems merely through regularisation methods or more ecologically plausible training regimes. However, we show that these problems can be ameliorated by forgoing end-to-end learning and processing images with Gabor filters in a manner that more closely resembles biological vision systems. These results raise doubts over the assumption that simply learning end-to-end in "vanilla" CNNs leads to the emergence of similar representations to those observed in biological vision systems. By adding more biological input constraints, we show that deep learning models can not only capture more aspects of human visual perception, but also become more robust to idiosyncratic biases within training sets.
Click here to insert in your Google calendar