Summary

Progress in computers and algorithms in the last years has made numerical simulation and modelling a key research methodology in both academia and industry, which in turn drives exascale computing in order to maintain excellence in research and innovation. A disruptive evolution in computer technologies is required for attaining exascale performances in the coming years bringing challenges that urgently need to be addressed across science and engineering fields. Therefore, new interdisciplinary strategies are required in order to educate the next generation of scientists to address such challenges enabling them to be at the forefront of their respective research fields. Instead of the traditional domain-specific training, integrated approaches are needed that can be best implemented by collaborative networks of universities, research institutes and industrial partners. HPC-LEAP is a highly interdisciplinary joint doctorate program realized by bringing together world-leading experts in applied mathematics, high performance computing technologies, particle and nuclear physics, fluid dynamics and life sciences to appropriately train researchers in Europe to exploit high performance computing, advance science and promote innovation. Students will be trained in mathematical and computational concepts underpinning current and future numerical simulations in turbulent flows, computational biology and lattice Quantum Chromodynamics. The research projects are designed to enhance collaborations and interactions across these disciplines, integrating non-academic partners, and to develop methodologies that efficiently use large-scale numerical simulations on future high performance computer systems. Students who complete this training program will be versatile to undertake highly interdisciplinary projects, well positioned to embark on a successful career in academia or the industrial sector.

Main Objectives

    • Provide a rigorous research program, which is interdisciplinary and at the same time has the required depth, by engaging world-leaders in computational science domains in a collaborative environment.
    • Contribute to developing a community of computational scientists in Europe that use simulation as a ‘third mode of discovery’, bridging experimentation and theory, by designing innovative HPC-driven research projects and training courses.
    • Advance the computational capabilities of the different communities by accommodating future HPC architectures requiring higher levels of parallelism and educate scientists to understand the challenges involved in the co-design approach and in managing big data associated with HPC.
    • Pioneer future, innovative HPC-related industries by integrating non-academic partners in the research projects effectively training young scientists in innovation practices.

Overview

HPC-LEAP consists of 15 interdisciplinary research topics defined by leading experts in scientific areas, which have in common the need for high-end supercomputing resources. Each application domain brings important training elements: Turbulent flows involve extremely diversified algorithms ranging from pseudo-spectral methods to kinematical approaches based on Lattice Boltzmann equations; Lattice Quantum Chromodynamics (QCD) has developed advanced Monte Carlo (MC) methods and involves linear solvers that are simple to program and easily portable on new computer architectures serving as a prototype application; Computational biology involves multilevel modelling and methodologies to integrate data especially in larger genomics projects and meta-analyses with the domain of infectious diseases permitting us to address some of these key issues. Integrating these topics with HPC architectures and technologies and Modelling and algorithms provides an ideal research environment to rigorously train young scientists in computational science providing human capital urgently needed for addressing global issues, such as personalized medicine and safer aircraft, as well as, furthering our theoretical understanding of turbulence and unravelling the structure of matter providing insight to the early universe. The program will award double degrees agreed between seven institutions in four EU countries while six additional academic and four non-academic partners will complement the training.

Fellow's individual projects*

*please note that titles are indicative

Research methodology and approach

The project is structured into 8 work packages (WPs). WP1 to WP5 are research WPs, while WP6 to WP8 concern management, training and dissemination. The project defines collaborative research projects in cross-disciplinary areas embedding them in three thematic domains. Early Stage Researchers (ESRs) are assigned to projects linking different WPs. The vast majority (11 of 15) link a thematic with a cross-disciplinary WP aiming to have training in HPC technologies and algorithms that will advance, on the one hand, the scientific disciplines and, on the other, the co-design approach. ESRs also work across thematic or cross-disciplinary WPs. This is essential in order to directly transfer novel methods from one domain specific field to another. The links among ESR projects and WPs are illustrated in Figure 1.

Fig1.: WP1-WP5: Research work packages and assignment of ESRs

Training programme

The training provided to the ESRs will be based on three main components: i) a number of core and elective courses complemented by shared lectures and seminars; ii) direct supervision through the experts in the network; iii) specialized training via secondments to academic and industrial partners. The role of these three components will be elaborated in the personal Career Development Plan (CDP) for each ESR. All the participating universities have well-established domain specific postgraduate programs. The core courses will complement those programs while the elective courses will enrich and improve the curriculum and student’s options at a given host. Students in the EJD programme 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 exception of TU/e for which it will be four years.

Core and elective 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 to be followed later on by thematically focused courses. This will provide a well-balanced mixture of general knowledge on HPC architectures and algorithmic design, and domain specific depth. Jointly with the research activities, the programme aims to produce graduates who are:

  • Versatile with the mathematical concepts underpinning current and future parallel numerical simulations;
  • Skilled software developers taking advantage of large-scale computing resources to advance their research;
  • Knowledgeable about the best data analysis techniques to achieve their research aims; • Confident in presenting their research findings to their peers in seminars and journal articles;
  • Flexible researchers, able to work on new projects in other disciplines, with a clear overview of contemporary academic and industrial research that uses large-scale numerical simulation;
  • Excellent in their chosen research discipline, with the Ph.D. qualification as evidence. At the successful completion of the program the ESRs will gain a double degree.

Content Structure

The courses offered in the program are divided into two groups, Group A is a set of core modules, while Group B consists of specialised modules for the thematic areas. An additional training component are the secondments to the industrial partners where10 ECTS will be given. Group A consists of two courses: CoS-1 Numerical analysis, algorithms and Monte Carlo methods (10 ECTS) and CoS-2 HPC architectures and programming (10 ECTS). Each core course includes two network-wide 3-week workshops. The workshops will be supplemented with additional exercises and course work that is completed at the nodes. Sharing of exercise material across nodes as well as lectures will be promoted and put on the training portal. Group B are domain-specific courses that complement the two Group A courses, the number of which will depend on the course requirements at each of the degree awarding institutions. Examples of Group B specialised modules that will be offered are: Computational biology and bioinformatics; Computational fluid dynamics; Quantum field theory; Molecular dynamics; Data mining and visualisation; Lattice quantum field theory; Numerical computing on graphical processing units. They can be taken at either degree awarding institution or any other institutions participating via secondments. These courses will be taught in English and can include seminars. These seminars will be streamed across partners and made available on the CyI training portal. In addition, three domain specific 1-week workshops will be organised. The ESRs must also attend at least one of these workshops and will be credited with 2 ECTS. Courses in Group B will be credited with the ECTS units applicable at each institution.

Network wide training events

Four cross-disciplinary and one domain specific workshop (listed in Table 1) will be compulsory for all students in the 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 with priority given to ESRs in cases where numbers are limited. The content and schedule of the workshops is as follows:

  • Month 6: School on Numerical analysis and algorithms towards exascale (part of CoS-1) – WP2: Introductory workshop presenting the algorithms and their mathematical foundations with emphasis on concepts for exascale computing. Topics covered will include Markov chain Monte Carlo algorithms, Molecular Dynamics, iterative solvers for linear systems and further methods from numerical analysis. Lead and Venue: BUW.
  • Month 8: School on HPC architectures and large-scale numerical computation techniques (part of CoS-2) – WP1: Topics covered will include HPC architectures and performance modelling, performance analysis and optimization, Parallel programming, data analysis, workflows and GPU programming concepts. Lead and Venue: JUELICH.
  • Month 11: Workshop on Numerical analysis and Algorithms at the exascale (part of CoS-1) – WP2: A continuation of the first school that will further advance concepts for designing scalable algorithms for particle-based simulation methods for CFD like LBM, multigrid solvers, symplectic integrators with multiple time-scales relevant for biological systems. Lead and Venue: RWTH.
  • Month 14: Workshop on HPC architectures and programming (part of CoS-2) – WP1: This will be a continuation of the second school and will include programming strategies on new computer architectures focussing on selected codes from the research areas of the project. Lead and Venue: TCD.
  • Months 19-22: Three topical workshops: Students will attend the most appropriate one for their research project. Turbulent flows – WP3, Lead and Venue: UTOV; Lattice QCD – WP4, Lead and Venue: DESY; Computational biology/infectious diseases – WP5. Lead and Venue: CyI. Each workshop will include a session introducing the methodologies of the other two domains.
  • Conference on research output (Month 39) – WP8: A conference emphasising the research outputs of ESRs as well as the insights gained by all research teams will be organised to showcase new ideas from the network. It will be open to the wider scientific communities. Lead and Venue: UC During the last core and all thematic workshops a day devoted to transferable skills will be organised. This will include a training day in presentation skills, a seminar on European patent law (we will aim to include a virtual classroom session from the European Patent Office, www.epo.org), a basic project management course (to be organised by the non-academic partners), a course in scientific writing and a basic proposal writing course (both organised by academic partners). Exposure of the majority of the ESRs (10 of 15) to the industrial innovation approach will occur during the 36 PMs non-academic secondments.
  Main Training Events & Conferences Venue Dates
1 3 week workshop – Numerical analysis and algorithms towards exascale

Wuppertal

23rd September-13th October 2015
2 3 week-workshop – HPC architectures and large scale numerical computation

Jülich 

11th - 29th  January 2016 
3 3 week workshop – Numerical Analysis and algorithms at the exascale RWTH Aachen

5th - 22nd April 2016 
4 3 week workshop – HPC architectures and programming

Trinity College Dublin 

13th June - 1st July 2016 
5 1 week workshop – Turbulent flows UTOV Project Month 19
6 1 week workshop – Lattice QCD DESY Project Month 20
7 1 week workshop – Computational biology/infectious diseases CyI Project Month 22
8 Conference UC Project Month 39

Table 1: Main Network-Wide Training Events, Conferences and Contribution of Beneficiaries

Supervision

Qualifications and experience of supervisors

We distinguish into supervisors from the degree awarding beneficiaries, and mentors from partners. All have supervised Ph.D. students in their respective research areas, the majority having over 10 years of experience. They collectively bring excellence in all areas of the project and the scientific expertise to successfully guide all ESRs. The eleven supervisors who will guide the ESRs are: C. Alexandrou, G. K. Christophides, G. Koutsou (CyI); M. Ehrhardt, A. Frommer, F. Knechtli (BUW); P. Carloni (RWTH); F. Toschi (TU/e); C. Alexandrou, H. Panagopoulos (UCY); R. Tripiccione (UNIFE); L. Biferale (UTOV). They will coordinate with the following twelve mentors during secondments: K. Jansen (DESY); G. K. Christophides, D. Vlachou (ICL); D. Pleiter (JUELICH); K. Iatrou (NCSR); M. Peardon, J. Bulava (TCD); M. B. Wingate, C. E. Thomas (UC), G. Tecchiolli (EUROTECH), A. Curioni (IBM); S. Kraemer (NVIDIA); J. Kallarackal (OakLabs).  In addition, Prof. U. Roethlisberger of EPFL agreed to provide expertise for ESR3 and will visit RWTH. Each of the above supervisors and mentors has a team of researchers who will be helping with the supervision.

Proposed joint supervision arrangements

The student will be based in two institutions, designated “Institution-A” and “Institution-B”, as listed in Table 2, with two supervisors, one from each host, jointly guiding the ESR. The students will normally, but not necessarily, follow a standard pattern where they commence their studies in Institution-A, which is also the recruiting institution, spend a period gaining new expertise in Institution-B and then return to Institution-A for the concluding period of their research while preparing their thesis. It is envisaged that the students evenly split their time between the two-degree awarding institutions. A third mentor from a node where the ESR has secondments will be assigned and together with the two supervisors will form the Supervision Committee (SC) that will monitor the students’ progress. The SC together with the ESR will produce a personal CDP at the beginning of the program that will define the objectives of the research and training and the stages for their achievement and enable effective progress monitoring. Meeting with a given supervisor will be frequent (typically weekly) and independent from geographic locations (use of VoIP).

Researcher No. Institution A
(Recruiting Participant)		 Institution B
(Second degree awarding Institution)
1. UNIFE BUW
2. CyI BUW
3. RWTH CyI
4. UTOV TU/e
5. UCY BUW
6. RWTH CyI
7. TU/e UTOV
8. UTOV CyI
9. BUW UTOV
10. CyI RWTH
11. UCY BUW
12. UCY BUW
13. BUW UCY
14. CyI RWTH
15. BUW CyI

Table 2: Recruitment Deliverables per Participant