The training provided to the ESRs will be based on three main components: i) five compulsory courses that include three-week network-wide workshops, complemented by lectures/seminars at the nodes that can be shared via video-streaming across institutions; ii) direct supervision through the experts in the network; iii) specialized training via secondments to academic and industrial partners. Progress will be tracked based on a Personal Career Development Plan (CDP) that is created at the beginning of the project. All the participating academic institutions have well-established domain specific postgraduate programs. Students in the STIMULATE project will be embedded in existing training structures and will benefit from interactions with other research students at the hosting and seconded institutions. The standard duration of the Ph.D. program will be three years. With the successful completion of the program the ESRs will gain a joint degree from three degree awarding institutions. Their degree will highlight their specialization in computational science via specially crafted Thesis fields stipulated in the detailed agreement among the three degree awarding institutions.
The compulsory (core) courses are designed to directly support the cross-disciplinary and thematic aspects of the training of the ESRs. Training will start on cross-disciplinary topics to give all fellows a uniform computational background. The four cross-disciplinary workshops that will form part of the core courses will take place during the first academic year of the ESRs in order to prepare them for their research projects. The last workshop associated with the fifth core course concerns applications of these tools to specialized problems in the research domains of focus and it is meant to demonstrate the applicability of the cross-disciplinary methodology across domains. Its aim is to enhance further the cross-fertilization and teach the ESRs to communicate across disciplines. It will take place within the second academic year. The research program opens possibilities of implementing innovative teaching methods such as reverse learning and sharing of courses among several institutions that will promote the establishment of a virtual European classroom. Some nodes (CyI, FZJ) already have experience in hosting HPC training content online such as the supercomputing portal hosted at CyI that combines audio-visual training material with embedded tutorials that run on a hybrid CPU/GPU cluster. The consortium pledges to record the training material of the network-wide courses in order to enrich the educational portal using the above tools.
All ESRs will be recruited at the same time and initially undergo an identical course program. The courses offered are to provide the cross-disciplinary fundamentals and give an overview of the methodologies in each specific domain. The research projects and complementary training at each institution will provide the required domain specific depth required. An additional training component are the secondments to the industrial partners where 10 ECTS will be given.
The courses offered in the program are divided into two groups, Group A is a set of core modules, while Group B consists of elective specialized courses or seminars.
The courses offered in Group A consist of a set of four core modules (8 ECTSs each) which include: CDS-1: High performance computing and simulation; CDS-2: Fundamentals of Data Science; CDS-3: Mathematical modeling and algorithms and CDS-4: Multiscale, multilevel algorithms and uncertainty quantification. These cross-disciplinary computational and data science (CDS) courses will be delivered during the first year and will provide the mathematical and computational background for carrying out the projects assigned to the ESRs. A fifth core course in Group A, CDS-5: Application to multiscale physical and biological systems, will focus on advanced methods in the three thematic domains of the project stressing the common aspects among them so that students learn to apply methods across disciplines. It will take place during the second year completing the highly innovative course requirements for the students.
Group B includes additional specialized courses or seminars that will be delivered by the local degree awarding institution. Examples of Group B specialized modules taught in English available on the various nodes are: Computational biology and bioinformatics; CFD; Quantum field theory; Molecular dynamics; Data mining and visualization; Lattice quantum field theory; Numerical computing on GPUs, etc. They can be taken at any institutions of the project. Courses in Group B will be credited with the ECTS units applicable at the institution, which offers them. All courses will consist of a network-wide workshop delivering two weeks of lectures accompanied with a project and supplemented with additional course work (e.g. in form of seminars). Exercise material as well as lectures/seminars at the nodes will be shared and be put on the training portal. The companies involved in the project will provide lectures on the skill needs for industry during the later three workshops.
Network-wide training events
The five compulsory workshops are part of the five core courses and are mandatory for all fellows. The respective hosting institution will take care of the local organization. A local committee will be set-up for each workshop that will always include the WP5 leader to ensure uniformity across the workshops and chaired by the scientist in charge at the local node. The local committee for CDS-5 will be chaired by the WP2 leader and also include the WP3 and WP4 leaders. The local committee will be responsible for delivering the workshop program. The workshops material will form part of the educational portal at CyI for future usage. These training events will be open, besides to the ESRs, to all students of the participating nodes as well as from outside the network. A summary of the workshop content is as follows:
- Month 7: High performance computing and simulation (part of CDS-1) – WP1, Lead/Venue: FZJ. The course will teach both theoretical as well as practical aspects of HPC. It includes lectures on performance architectures and modeling as well as lectures related to different skills that are required for using HPC systems efficiently. This includes, e.g. parallel programming, GPU programming, performance engineering etc. The lectures will be complemented by keynotes given by computational scientists that have experience in using highly scalable systems for their research. A project will be carried out at the venue to take advantage of the exascale labs.
- Month 9: Fundamentals of Data Science (part of CDS-2) – WP1, Lead/Venue: UNIFE. This workshop will introduce students to state-of-the-art software methodologies and technologies to handle, manage and analyze the large amounts of data coming from extensive simulations using innovative algorithms. It will cover three main areas of interest: data visualization in scientific contexts, statistical methods for data-analysis, machine-learning and deep-learning with special focus on neural-networks. It will include lectures in which the mathematical principles and the relevant methodologies will be presented in detail, and hands-on training will be provided on the topics covered. A fraction of the workshop will be devoted to the development of projects assigned to each student on one of the three scientific areas covered including methods for dealing with model comparison.
- Month 11: Mathematical modeling and numerical analysis for exascale (part of CDS-3) – WP1, Lead/Venue: HUB. Topics covered will include partial differential equations, times series, introduction to iterative solvers for linear systems, Conjugate Gradient (CG), preconditioned CG, Fourier and polynomial spectral methods, design of scalable algorithms for particle-based simulation methods for CFD like Lattice Boltzmann methods (LBM), Hybrid Monte Carlo (HMC), symplectic integrators with multiple time-scale relevant for biological systems.
- Month 13: Multiscale, multilevel algorithms and uncertainty quantification (part of CDS-4) – WP1, Lead/Venue: BUW. Linear as well as nonlinear multilevel and multigrid methods including domain decomposition methods for linear elliptic problems, (recursive) trust-region methods and their application to minimization problems and their parallelization in view of exascale architectures.
- Month 20: Applications to multiscale physical and biological systems (part of CDS-5) – WP2, WP3, WP4, Lead/Venue: UTOV. The workshop will focus on data-driven multiscale problems arising in all the main topics of STIMULATE, with a particular emphasis on interdisciplinary aspects. We will touch short-distance and long-distance subnuclear physics for lattice QCD+QED simulations (including heavy particles, ultraviolet singularities, renormalization, decay and scattering amplitudes, finite-volume effects); deep learning for biological image analysis and regulatory genomics; statistical and optimization approaches to rare events; and finally a set of paradigmatic cases for flowing matter at the crossroads between industrial processes, fundamental physics and biology at nano-, micro- and macro-scales (wall slip, thin films, moving contact lines, emulsions and soft-glassy materials, particles in turbulent flows). The workshop is meant to expose young researchers to the most recent advancements in these fields, by dealing with theoretical and numerical studies.
- Month 40: A conference on Multiscale Physical and Biological Systems – WP6, Lead/Venue: CyI. A conference emphasizing the research outcome of the ESRs as well as the insights gained by all research teams will be organized to showcase new ideas from the network. It will be open to the wider scientific communities.
|Main Training Events & Conferences
|3-week workshop on High performance computing and simulation
|3-week workshop on Fundamentals of Data Science
|3-week workshop on Mathematical modeling and numerical analysis for exascale
|3-week workshop on Multiscale, multilevel algorithms and uncertainty quantification
|3-week workshop on Applications to multiscale physical and biological systems
Training opportunities within the network
No single node offers such a broad range of opportunities. Computational Science (CoS) programs are offered at the Master’s level by BUW and RWTH and at the Ph.D. level by CyI, which also runs a dual degree with the University of Illinois at Urbana Champagne (UIUC). Graduate courses taught at each node will be open to ESRs adding competence to the thematic training across the network. Selected seminars will be streamed across the nodes thus enriching the scientific exposure of the fellows and leveraging the expertise at each institution. The ESRs will join very active research groups in their host institutions. These groups organize more informal research training, such as journal clubs and seminar programs. The EJD scholars will be expected to participate in these activities. As research progresses, the case might arise that particular expertise is available in another network node. Students will be enabled and encouraged to travel on secondment to these nodes as appropriate. Experts from or outside the network will also be invited to give specialized training if needed.
Role of non-academic sector in the training programme
The non-academic partners in the network add a vital industrial and commercial dimension through providing unique training “on-the-job” in an industry environment, exposure to commercial practices and priorities, providing the opportunity to develop relationships, leading to future employment opportunities and a transfer of knowledge framework between industrial and academic stakeholders. The non-academic partners have a scientific track record and will actively contribute to the training by hosting secondments. From the 15 ESR projects proposed here, 11 include secondments with the non-academic partner specifically selected to enrich the respective Ph.D. project. Besides secondments, the three companies, IBM, NVIDIA and MAGWEL, will participate in the training workshops and give specialized lectures as outlined above.