|
VAMPIREPrinciple Investigator: Richard Evanshttps://pure.york.ac.uk/portal/en/persons/richard-francis-llewelyn-evans Project DescriptionMagnetic materials are ubiquitous today, and essential to many technologies, from powering your electric motor to storing images and videos from Facebook. In this project, we will work on developing the next generation of advanced technologies based on magnetic materials, with applications in cancer treatment, wind power, data storage, and artificial intelligence. Harnessing the power of the Bede supercomputer will enable unprecedented calculations with atomic-scale resolution, allowing us to really understand the fundamental nature of advanced magnetic materials and devices.
|
|
Supervised and Self-Supervised Relative Pose Estimation for Planar ScenesPrinciple Investigator: William Smithhttps://pure.york.ac.uk/portal/en/persons/william-alfred-peter-smith Project DescriptionOur research explores a novel approach to training a deep neural network for the task of monocular visual odometry for road scenes (essential for tasks such as autonomous driving) by modelling the road as a planar surface. Our approach could help greatly simplify the training of state-of-the-art methods in novel locales to make real-time relative pose estimation achievable for smaller organisations and more accessible to wider audiences.Combined with road segmentation, we are implementing convolutional neural networks trained with a homographic appearance loss to learn local planar features of the road. A mixture of supervised and self-supervised appearance loss could allow us to learn efficient representations of a road scene, which potentially could allow for very efficient odometry when combined with GPS.
|
|
Understanding and Manipulating Human Memory for ImagesPrinciple Investigator: Karla K. Evanshttps://pure.york.ac.uk/portal/en/persons/karla-evans Project DescriptionThe project aims to both understand what makes images memorable for humans and to use this understanding to create new images that have a set memorability value. We employ a concept from cognitive psychology, that of the ‘Visual Memory Schema’, which defines the structural organization of a given image that an average person is likely to remember. These schemas are based upon learned cognitive structures, e.g., an image of a kitchen is more memorable because it matches our mental representation of a kitchen. These cognitive concepts are incorporated into state-of-the-art neural network models, aimed at both understanding and predicting these visual schemas, and also using them with advanced image generation techniques to generate brand new images that are as memorable for humans as we wish them to be.
|
|
First-principles materials modelling on pre-exascale HPCPrinciple Investigator: Phil Hasniphttps://pure.york.ac.uk/portal/en/persons/philip-james-hasnip Project Description"First-principles materials modelling" uses quantum mechanics to predict the properties of new and existing materials. These computer simulations have become a cornerstone of materials science, aiding in the interpretation of experimental data and guiding experimental design, and have benefited enormously from a substantial year-on-year increase in available computing power. In recent years, however, the increases in computer power have come about through new kinds of computers, many of which use GPUs to perform the most intensive calculations--Bede is one such machine.Using GPUs effectively requires different techniques to conventional CPUs, and in this project, the CASTEP first-principles materials modelling code will be adapted to use Bede's GPUs much more efficiently. This will dramatically improve the speed at which scientists can perform materials simulations on Bede, in turn leading to more and greater materials discoveries.
|
|
Metagenome sequence polishingPrinciple Investigator: James Chonghttps://pure.york.ac.uk/portal/en/persons/james-chong Project DescriptionWe are interested in the dynamics that occur in complex microbial communities such as those found in soil and the human gut. These "microbiomes" can be explored using DNA sequencing. Innovations in DNA sequencing technologies allow the low-cost generation of very large datasets. We want to investigate how the power of GPU systems can be used to improve the quality of the DNA sequences we can generate using new technologies, and explore approaches that reduce the time required to process our datasets.
|
|
cryo-EM of respiratory complexesPrinciple Investigator: Jamie Blazahttps://pure.york.ac.uk/portal/en/persons/jamie-blaza Project DescriptionWe use electron microscopes to image the machines that provide pathogenic bacteria, like tuberculosis, with energy. These processes allow us to understand how existing drugs work and are also validated drug targets. For example, the new antibiotic against tuberculosis, bedaquiline, has recently been approved and is a key weapon in our arsenal against drug-resistant bacteria. In the microscope, we image thousands or even millions of these enzymes and computationally use the 2D images to create a 3D 'map' of the tiny machines so we can understand how they work and how they might be targeted by new antibiotics. This process is incredibly computationally expensive, so we use the specialised graphics-focused cards at Bede to enable our work.
|
|
Artificial Intelligence-Driven Resource Allocation Techniques for Future NOMA Systems.Principle Investigator: Abdulhamed Waraiethttps://pure.york.ac.uk/portal/en/persons/abdulhamed-khaled-e-waraiet/ Project DescriptionHand-crafted convex optimization algorithms have been used as the only methods for solving optimization problems in wireless communications. However, such a method suffers from a major design flaw called stationarity; even a slight change in the model assumptions may render it useless or severely degrade its performance. Data-driven models generated by applying machine learning algorithms have received great attention in research communities across different disciplines in recent years. In this project, we aim to develop novel deep reinforcement learning methods to solve resource allocation problems in future wireless networks and provide a better alternative to hand-crafted methods.
|
|
Modelling DNA in complex 3D arrangementsPrinciple Investigator: Agnes Noyhttps://pure.york.ac.uk/portal/en/persons/agnes-noy Project DescriptionDNA is the molecule that nature uses as genetic material, and it rarely exists in a relaxed state. Rather, it is subjected to torsional, bending or stretching stress generated in cellular processes such as recombination, gene expression, replication, and more generally protein recognition. These tensions cause severe distortions on the double helix, which, in turn, influence its dynamic and recognition properties. Our group is focused on modelling DNA fragments longer than 300 bp at atomic resolution under mechanical stress, which causes the formation of complex spatial arrangements. This nano length-scale is the missing link between crystallographic structures, which usually contain short DNA fragments (up to 20 bp), and force-extension experiments, which are applied to DNA molecules of some kbps. The use of the Bede supercomputer has enabled us to study stretched, bent, and supercoiled DNA, together with its melting caused by the action of cellular motors. All these aspects of the molecule of DNA are relevant for understanding its functionality inside living beings.
|
|
Structure, mechanism and dynamics of recoding in viral infectionPrinciple Investigator: Chris Hillhttps://pure.york.ac.uk/portal/en/persons/chris-hill Project DescriptionAccurate translation of mRNA by the ribosome is essential for all life. It is therefore a high-fidelity process, with spontaneous error rates of only ~ 1 in 100,000 codons. When RNA viruses infect cells, they frequently make proteins that differ from the genetically encoded sequences. These highly-regulated ‘recoding’ events are vitally important to viral gene expression, and if disrupte,d many viruses (e.g. SARS-CoV-2, HIV-1) fail to complete their replication cycles. Our group uses a multidisciplinary approach to study recoding. We combine single-molecule fluorescence microscopy, time-resolved cryo-EM, X-ray crystallography, and biophysical approaches to reveal the universal structural and mechanistic principles involved. A better understanding of these processes will lay the foundation for a new class of antiviral drugs.
|
|
Embedding FAIRness in Plasma SciencePrinciple Investigator: Liam Pattinsonhttps://plasmafair.github.io/ Project DescriptionThe PlasmaFAIR project aims to support the sustainability and reusablility of data and software within plasma physics by implementing the FAIR principles, which state that data/software should be Findable, Accessible, Interoperable, and Reusable. We offer free software health checks and short software engineering projects to all researchers within plasma science. Our work often includes packaging research software for wider distribution, writing automated testing suites and user guides, and implementing software features that promote usabilty, robustness, and sustainability. For more information, please see https://plasmafair.github.io/.
|
|
Tailoring health policies to improve outcomes using machine learning, causal inference and operations research methodsPrinciple Investigator: Julia Hatamyarhttps://pure.york.ac.uk/portal/en/persons/julia-farideh-hatamyar Project DescriptionIn this research project, I am actively engaged in using causal machine learning techniques to investigate the impacts of health policy on health outcomes in low and middle-income countries. My primary focus revolves around developing models that evaluate the heterogeneous effects of primary healthcare access and health insurance subsidies on infant and maternal mortality. I also explore ways of learning how to deploy limited health resources optimally - i.e. to those groups who need them most. Additionally, I contribute to the advancement of methodology by working on the intersection of econometrics and machine learning, creating algorithms to enhance our understanding of how different groups of people experience different benefits from health insurance at different times.
|
|
Engineering proteins for applications in biosensorsPrinciple Investigator: Jake Kerrisonhttps://www.york.ac.uk/biology/people/michael-plevin/ Project DescriptionThere is an important need to sense for small molecules in solutions. Proteins are a medium to design binding sites tailored to particular small molecules, so they can bind small amounts selectively. To effectively design these protein binders, molecular dynamics simulations need to be run to understand the flexibility and conformations of both the target and receptor to guide the design trajectory.
|
|
The York Physics of Pyrenoids Project: Nanostructured Biological LLPSPrinciple Investigator: Clement Deguthttps://pure.york.ac.uk/portal/en/persons/clement-degut Project DescriptionSingle-cell algae are responsible for 30% of global carbon fixation. This process is aided by a structure unique to these cells: the pyrenoids. The York Physics of Pyrenoids Project is harnessing high-power computing infrastructure to help understand the complicated physical and biological characteristics of these membraneless organelles.
|
|
Turbulence and transport in the core of high-𝛽 spherical tokamaks and predictions for STEPPrinciple Investigator: Maurizio Giacominhttps://pure.york.ac.uk/portal/en/persons/maurizio-giacomin Project DescriptionThis project addresses the turbulent transport in the core of magnetic fusion devices and it is focused, in particular, on the evaluation of particle and heat fluxes in some proposed scenarios for the Spherical Tokamak for Energy Production (STEP) project. STEP is an ambitious UK project that aims to demonstrate the feasibility of producing net electricity from fusion energy, which is one of the most promising alternatives to fossil fuels. The design of future magnetic confinement fusion power plants requires the accurate prediction of several quantities. Among these, the energy confinement time, which depends on turbulent transport, is extremely important. First-principle evaluation of particle and heat transport requires complex simulations. In this project, simulations are carried out by using the GX code, which is a code specifically designed to natively target GPUs, thus enabling fast simulations at reactor scale.
|
|
Artificial Intelligence-Driven Resource Allocation Techniques for Future Wireless NetworksPrinciple Investigator: Kanapathippillai Cumananhttps://pure.york.ac.uk/portal/en/persons/kanapathippillai-cumanan Project DescriptionCell-free massive MIMO (CF-MaMIMO) is a potential candidate for enabling real-world artificial intelligent (AI)-driven use-cases, such as autonomous vehicles. Inheriting advantages from other types of MIMO, CF-MaMIMO offers excellent user experience, higher data rates, and massive connectivity. Despite these advantages, existing resource allocation techniques for CF-MaMIMO are not scalable for future demands. They completely rely on conventional iterative optimizations without using gathered data. Thus, AI-driven techniques for CF-MaMIMO, in the current climate of growing on-device capabilities, are the logical solution. This project will develop novel AI-driven resource allocation techniques for CF-MaMIMO. Fundamentally, we will use AI to accelerate the convergence of the iterative frameworks. Strategically combining both AI and optimization techniques, the developed scalable approaches will enable CF-MaMIMO to support envisioned use cases.
|
|
Exploratory work to process cryo-EM test datasets using RelionPrinciple Investigator: Johan Turkenburghttps://www.york.ac.uk/chemistry/research/ysbl/facilities/eleanor-and-guy-dodson-building/ Project DescriptionIn recent years, the University of York has established a state-of-the-art facility for Structural Biology in a purpose-built laboratory, bringing together cryo-EM, X-ray crystallography, and NMR. To provide optimal settings and alignment of the instruments test datasets are collected in known proteins and these are processed using various available platforms, including Bede to facilitate specialist support for researchers taking up use of our facilities.
|