This proposal led by the University of Oxford, with support from the Alan Turing Institute (ATI), Bristol, Edinburgh, KCL, QMUL, Sheffield, Southampton and UCL is for a national GPU system that will support multidisciplinary science with a focus on machine learning and molecular dynamics. The architecture is based on ``fat'' GPU compute nodes, with 8 of NVIDIA's new Pascal GPUs, each with a) 16GB 720GB/s HBM2 memory, b) an 80GB/s NVlink interconnect to other GPUs, c) 6GB/s bandwidth to main system memory, d) 6GB/s bandwidth to the Infiniband external network. Each server also has two 20-core Xeons, 512 GB DDR4 memory and 8TB SSD. The motivation for selecting this architecture is the huge growth in research in machine learning and associated areas of data science within the UK, particularly within the universities which are members of the Alan Turing Institute, or SES. The same architecture is also ideally suited for molecular dynamics, medical imaging and a number of other application areas. The system will be run as a national facility, similar to Archer in being free to all academic users with computing time available to all through a lightweight Resource Allocation Panel, with a top-level steering committee determining the policy on resource allocation between the different application areas (Machine Learning, Molecular Dynamics, Other).

Central to many research programmes in plasma physics is the requirement to capture the full kinetic properties of the plasma. This is most commonly done with a particle-in-cell (PIC) code which move samples of plasma particles in the self-consistent electromagnetic field. For over a decade Warwick, initially in collaboration with Oxford and York, has been developing such a code for UK plasma researchers - the EPOCH code. The core PIC scheme deals with a plasma in which collisions between particles are ignores - a collisionless plasma. EPOCH has been extended to include binary collisions, Quantum Electro-Dynamics (QED) processes and radiation. This has made EPOCH one of the most widely used PIC codes worldwide with application to particle accelerators, next generation light sources, high-power laser QED studies and fusion science for both magnetic and inertial confinement fusion. There are key science extensions required to UK EPOCH to meet the needs of the UK user base. These are detailed in the Objectives section above. However, maintaining EPOCH has become increasingly fraught. This is because EPOCH evolved out of an older Fortran PIC code and while this was relatively easy to modify for the collisionless plasma when adding more sophisticated boundaries, diagnostic or physics packages the restricted feature set of Fortran made these additions increasingly cumbersome. It may be possible to persevere with the old legacy Fortran for a while longer but there are other perhaps more pressing problems. Modern high-performance computer (HPC) systems are moving away from being built all from the same processor type. Instead HPC systems with a mixture of traditional CPUs accelerated with GPUs are becoming prevalent. Other systems are being installed which use ARM processors or even FPGAs. This move towards heterogeneous HPC systems and a complex HPC landscape is been driven by the need for continued increases in computing power. The toolsets for working with these newer heterogeneous systems are primarily in C++, not Fortran. Moving EPOCH over to a C++ implementation (renamed EPOC++ as a result) opens up access to these newer routes towards Exascale HPC. At the same time it makes the code easier to maintain and more flexible for future changes. This project will complete all these changes and deliver an EPOC++ code able to meet the needs of UK and International EPOCH users with a performant and parallel code future proofed for decades.

Mathematical competence can endow robots of the necessary capability for abstract and symbolic processing, which is required for improving their cognitive performance and their social interaction with human beings. But so far, only few attempts have been made to apply mathematical cognition in robots. The objective of the NUMBERS project is to construct a novel artificial cognitive model of mathematical cognition by imitating human like learning approaches for developing number understanding. This will provide a novel tool for developmental psychology and neuroscience research on the development of mathematical abilities in children. The objective will be achieved through a highly interdisciplinary research program that will take advantage of the collaboration of leading academics in the fields involved, prof. Cangelosi (Cognitive Developmental Robotics), Plymouth University, and prof. McClelland (Computational Psychology), Stanford University, USA, and the technical support of an industrial partner (NVIDIA Corporation, USA and UK). The NUMBERS' research activities will exploit the cutting-edge facilities offered by Sheffield Robotics, a joint initiative between the University of Sheffield and Sheffield Hallam University, which will host the project. Indeed, the model will be integrated with one of the most advanced child-like robotic platform (the "iCub") and, therefore, validated through realistic experiments, resembling scientific experiments of mathematical cognition in children. The validation of the novel model of the numerical cognition in interactive robotic experiments will constitute a proof of concept of the enhanced capabilities offered by a modular approach to bio-inspired artificial intelligence architectures. Furthermore, an optimised implementation for mobile devices will be realised in order to downsize space and power requirements for the computation, increasing application opportunities. This foundational research will provide the methodological basis and cognitively plausible engineering principles for the next generation of socially interactive robots, mimicking advanced capabilities of the human intelligence for real understanding and interaction with the external world. Results will help the design of more efficient cognitive robotic systems capable of learning abstract symbolic number processing in a more flexible and ecological manner. The human-like learning and interaction are characteristics that might allow people to more easily identify the desired social overture that the robot is making, or facilitate the transfer of skills learned in human-human interactions to human-robot encounters. This envisioned humanization will positively affect the acceptance of robots in social environments, as they will be perceived as less dangerous, increasing the socio-economic applications of future robots that can take on tasks once thought too delicate or uneconomical to automate. This is particularly relevant in the fields of social care, companionship, therapy, domestic assistance, entertainment, and education.

The equipment requested will provide new capability and internationally leading facilities that will enable cutting-edge research and internationally leading science, beyond that which is possible with current instrumentation. The equipment will also facilitate greater collaborative opportunities with other Universities and industry nationally and internationally. The "Advanced Electronic Materials and Devices" bundle provides equipment for research into new materials and devices for future electronic applications, ranging from superconductors for applications in power transmission and MRI to spintronic devices for sensors and computer memory applications. It will also improve thermal imaging capability for the study of semiconductor and hybrid diamond based devices which have the potential to transform future power electronic devices. Electrical power conversion technologies have a vital role to play in managing energy demand and improving energy conversion efficiency, affording 'game-changes' in, for example, low carbon transport systems and energy supply networks. As these 'more electric' systems become more commonplace, for example through their adoption in aircraft and electric vehicles, new understanding of operation life and failure modes is needed. The enhanced capabilities offered by the equipment updates in the "Enabling robust design and analysis of electrical power conversion systems" will allow internationally leading research to be pursued in the areas of design for life, virtual certification and reliability. Transmission electron microscopes (TEM) allow the imaging of both the external and internal structure of materials and are available in numerous configurations dependent on the specific nature of the materials under investigation. A post column energy filter dramatically improves the analytical and imaging capabilities of a TEM by allowing structural and chemical information carried by the electrons to be interrogated after interaction with the sample material. The requested Gatan Imaging Filter (GIF) upgrade in the "Supporting Analysis of Advanced Energy Materials and Soft Matter" will provide significant new capability to determine the structure and composition of materials at the nanoscale and provide new insights into how to enhance material functionality. The instrument upgrade forms part of a strategic investment in advanced microscopy provision at Bristol, and reflects ambitions for an internationally competitive materials characterization facility befitting the world-leading research it underpins. Nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS) are the pre-eminent techniques for studying chemical structure and reactions. They underpin nearly every program of research in catalysis (accelerating chemical reactions), synthesis (creation of new chemical entities) and materials (chemicals with defined properties and applications e.g. nanotechnology). The replacement of aging NMR and MS instruments as described in the "Underpinning Catalysis, Synthesis and Materials Chemistry" bundle will ensure continued cutting-edge investigations in these fields, and will provide new hardware capabilities that allow the study of molecular/chemical systems in previously impossible fashions, e.g., at low temperature for days at a time (NMR), or under unreactive atmospheres (MS). The new "Wideband Multi-channel Real-time Wireless Channel Emulator" facility will offer wideband (160MHz) multi-dimensional channel (8 x 8) wireless channel emulation for sub 6GHz wireless transceivers allowing repeatable experimentation with real-world channel models (3GPP and 802.11, plus user defined scenarios). The hardware can also be reconfigured to offer channel emulation with cascaded bandwidths synonymous with millimetre wave operation, thus driving forward the "5G and beyond" research agenda.

We are living through a revolution, as electronic communications become ever more ubiquitous in our daily lives. The use of mobile and smart phone technology is becoming increasingly universal, with applications beyond voice communications including access to social and business data, entertainment through live and more immersive video streaming and distributed processing and storage of information through high performance data centres and the cloud. All of this needs to be achieved with high levels of reliability, flexibility and at low cost, and solutions need to integrate developments in theoretical algorithms, optimization of software and ongoing advances in hardware performance. These trends will continue to shape our future. By 2020 it is predicted that the number of network-connected devices will reach 1000 times the world's population: there will be 7 trillion connected devices for 7 billion people. This will result in 1.3 zettabytes of global internet traffic by 2016 (with over 80% of this being due to video), requiring a 27% increase in energy consumption by telecommunications networks. The UK's excellence in communications has been a focal point for inward investment for many years - already this sector has a value of £82Bn a year to the UK economy (~5.7% GDP). However this strength is threatened by an age imbalance in the workforce and a shortage of highly skilled researchers. Our CDT will bridge this skills gap, by training the next generation of researchers, who can ensure that the UK remains at the heart of the worldwide communications industry, providing a much needed growth dividend for our economy. It will be guided by the commercial imperatives from our industry partners, and motivated by application drivers in future cities, transport, e-health, homeland security and entertainment. The expansion of the UK internet business is fuelled by innovative product development in optical transport mechanisms, wireless enabled technologies and efficient data representations. It is thus essential that communications practitioners of the future have an overall system perspective, bridging the gaps between hardware and software, wireless and wired communications, and application drivers and network constraints. While communications technology is the enabler, it is humans that are the producers, consumers and beneficiaries in terms of its broader applications. Our programme will thus focus on the challenges within and the interactions between the key domains of People, Power and Performance. Over three cohorts, the new CDT will build on Bristol's core expertise in Efficient Systems and Enabling Technologies to engineer novel solutions, offering enhanced performance, lower cost and reduced environmental impact. We will train our students in the mathematical fundamentals which underpin modern communication systems and deliver both human and technological solutions for the communication systems landscape of the future. In summary, Future Communications 2 will produce a new type of PhD graduate: one who is intellectually leading, creative, mathematically rigorous and who understands the commercial implications of his or her work - people who are the future technical leaders in the sector.