HetSys' training programme is to enable students to become high-quality computational scientists who are comfortable working in interdisciplinary environments, have excellent communication skills, and well prepared for a wide range of future careers in areas where there is demonstrable need.
The HetSys training programme will meet three key training needs:
- Span disciplinary barriers. The most challenging real-world heterogeneous systems are intrinsically multidisciplinary, requiring integration of a diverse range of modelling methods.
- Incorporate uncertainty in modelling. Training in uncertainty quantification will enable students not only to perform simulations, but also to quantitatively assess their reliability.
- Promote robust Research Software Engineering (RSE). Training in sustainable software development will enhance software usability and extend its lifetime.
Overview of Training Programme
CDT training will run throughout the four year PhD programme as illustrated below, with all aspects designed to meet the three key training needs above, and to develop transferable skills. Students will be recruited directly onto projects and will have a supervisor from the start of their course, as well as a second supervisor in a related area and a cohort mentor for academic and pastoral advice. In projects with industry links there will often also be an industry co-supervisor.
During the first 18 months of the programme each student will study 4 core modules (PX911, PX912, PX913 and PX914) and at least 2 optional modules, participate in a group software development project (PX915) supported by academics and RSEs, and carry out an independent research project in the area of their PhD project assessed through a written report and viva 12 months into the programme.
The individual project also leads to a peer-to-peer activity in the second year (also part of PX915). These activities will contribute to the formal award of a postgraduate diploma (120 credits), which must be successfully completed 18 months into the programme. Each student will also have the opportunity to participate in the formal transferable skills course run by the University, which leads to a PG Certificate in Transferable Skills after 3 years.
In Years 2-4 the majority of the students’ time will be spent conducting PhD research. There will be ample opportunities for peer-to-peer learning and knowledge exchange through cohort-wide activities.
HetSys Core Modules
PX911: Multiscale modelling methods and applications I will provide an introduction to atomistic modelling techniques including DFT, classical force field methods and an appreciation of how they interact with other modelling frameworks. Students will learn how to design atomistic simulations of condensed matter or molecular systems, and how to identify simulation methodologies appropriate to bridging multiple length scales, balancing accuracy vs. cost. They will gain exposure to software packages supporting interoperability between methods, e.g. the Atomic Simulation Environment. Multiscale Modelling case studies by guest lecturers will show how problems involving heterogeneous systems are tackled at multiple length & time scales.
PX912: Multiscale modelling methods and applications II will provide a firm grounding in macroscopic and multiscale modelling techniques, with lectures on foundations of continuum mechanics, thermodynamics, fluid dynamics, solid mechanics, and recent developments in multiscale fluid, plasma and solid mechanics, again with an emphasis on applications and on the links between methods and across scales. Topics covered will range from the basics of continuum mechanics and thermodynamics concepts through to demonstrating the route from underlying models via algorithms to practical implementation in simulation packages.
PX913: Introduction to Scientific Software Development will comprise bespoke Software Carpentry training developed by our RSE group and will ensure students understand the core principles of programming and software development, gain experience with writing, debugging and reading code in high- and low-level languages, and learn to use common tools for data analysis and visualization. Lectures will be delivered by the Research Software Engineering group of the Scientific Computing Research Technology Platform, and will cover fundamental operation of a computer, use of version control, debugging tools, and approaches to group-based software development. Where necessary for individual projects, C or Fortran training will be provided, positioning students to follow further programming option modules, most notably PX425: High Performance Computing.
PX914: Predictive Modelling and Uncertainty Quantification will give an introduction to predictive modelling techniques including statistics, machine learning, data analytics and data mining, essential for solving problems in the interdisciplinary area of predictive modelling. Lectures will cover random processes, statistical learning, Bayesian inference, Monte Carlo methods, model selection, and supervised and unsupervised machine learning techniques. Through links to topics in PX911 and PX912, students will learn how to quantify uncertainty in a range of modelling approaches. Particular emphasis will be given to scalable approaches to UQ and propagation in multiscale models, description of random microstructures, defects in random media and information theoretic approaches to coarse graining.
PX915 Part I: Software engineering group project, supervised by a combination of academics and RSEs, will be undertaken collectively by the cohort, with students designing, specifying, optimising and implementing a small-scale simulation package to model heterogeneous systems. Examples include density functional theory, molecular dynamics, computational fluid dynamics and finite element analysis. Careful division of responsibilities and integration of work will be required. This will be valuable for students’ future careers, as well as making it easier to distribute software developed in their PhD research. This 25 CAT module will include 2 seminars on intellectual property and software licensing with input from Warwick Ventures, who commercialise innovations produced from research carried out in the University.
Year 1 students will also complete a 12-week individual research project which requires detailed field-specific knowledge and approaches the frontiers of research, with the innovative requirement to include explicit quantification of uncertainties and/or modern aspects of software design. In general, the project will be preparation for the main PhD work. The first year report is not for credit within the PG Diploma, which allows the work undertaken to be included in the PhD thesis, but the supervisory team must be satisfied with the progress during this time for progression to the PhD.
PX915 Part II: Peer-to-peer Project Evaluation (Year 2) will build on output of individual research projects: another cohort member will be assigned to evaluate the quality of a student’s model error estimates and/or compliance with research software engineering principles, e.g. by running models with an ensemble of inputs or on a selection of architectures (another opportunity for PUE profiling). Software testers will be encouraged to discuss the process with the software authors, fostering dialogue across the cohort.
Across Years 1 and 2, students must choose at least 2 optional modules (totalling 30+ credits) from an approved set of MSc-level modules including 7 bespoke HetSys modules: these allow students to develop necessary theoretical background for their PhD projects, and gain hands-on experience with algorithms and software packages in their field.
HetSys Optional Modules
Students will be required to study 15 or more credits from optional modules that cover topics including:
- PX917 Computational Plasma Physics (15 credits)
- PX918 Electronic Structure Theory for Experiment and Models (15 credits)
- PX919 Quantum Chemistry (7.5 credits)
- PX920 Homogenisation of Nonlinear Heterogeneous Solids (7.5 credits)
- PX921 Micro & Nano Flows across scales (7.5 credits)
- PX922 Approximation theory for partial differential equations and machine learning (15 credits)
- PX923 Biomolecular Simulation (7.5 credits)
HetSys students will also have access to a range of other relevant postgraduate modules from across Warwick, for example:
- MA4K0 Introduction to Uncertainty Quantification (15 credits)
- PX425 High Performance Computing in Physics (7.5 credits)
- MA933 Stochastic Modelling and Random Processes (12 credits)
- MA934 Numerical Methods and Algorithms (12 credits)
- Fundamentals of Mathematical Modelling (12 credits; new module)
- MA930 Data Analysis and Machine Learning (12 credits)
- IB9FF Simulation Optimisation (15 credits)
- CS910 Foundations of Data Analytics
- CY905 Computational PDEs
- MA4L4 Mathematical Acoustics
- MA913 Scientific Computing