High Performance Computing for neuroscience:
Hands-on introduction to supercomputing usage, tools and applications

Anne Carstensen holds a Diploma in Molecular Biomedicine and a PhD in Molecular biology/Tumour biology focusing on Neuroblastoma and protein-protein interactions with the N-MYC oncogene. Formerly she worked as scientific coordinator at the University of Bonn and since 2018 she is a project manager at the Jülich Supercomputing Centre and a member of the “Simulation Laboratory Neuroscience”, working in ICEI (Fenix), HBP (Subproject 7)  and PPI4HPC.

Lecture title: Access to the Fenix Infrastructure

The talk focusses on providing information on how users can get access to resources of the Fenix Infrastructure. After a short introduction into the Fenix Infrastructure and its services, the participants will get to know which resources are currently available and how they can apply for resource allocations as an HBP member and via PRACE.

Benjamin Cumming is a computational scientist in the Software and Libraries group at the Swiss National Supercomputing Center. He has extensive experience in developing, optimizing and porting scientific software for different HPC architectures. Currently Benjamin Cumming is leading the development of Arbor, a performance portable library for simulation of large, distributed networks of multicompartment neurons under the aegis of the HBP.

Lecture title: Getting started with Arbor

Arbor is being developed by Subproject 7 (HPAC Platform) in the HBP for efficient simulation of multi-compartment models on different computing architectures, including GPUs and Intel KNL. This course will give an overview of Arbor, covering:

• Features and capabilities of Arbor;

• The design of Arbor;

• Target use cases suitable for simulation with Arbor.

There will be a practical demonstration of how to download, setup and use Arbor, with a focus on using the Python interface to run a model on an HPC system.

Fahad Khalid is a member of the “Simulation Laboratory Neuroscience” at the Jülich Supercomputing Center, where he leads the “Analysis, Visualization, and Learning” team. His current projects include applications of deep learning to neuroimaging, and development of tools for analysis and visualization of graph theoretic models in neuroscience. Fahad holds undergraduate and master’s degrees in computer science. He did four years of research in parallel programming, followed by roughly two years in computational evolutionary biology. In addition, Fahad spent five years in the software industry designing and developing large-scale GSM/3G telecommunication systems. 

Lecture title: Introduction to Python part I & II

This lecture will introduce the students to the basic concepts of the Python programming language: data types, control structures, exception handling, and the usage of modules. In addition, a short overview of the standard Python libraries will be given. It is expected that the participants have already some programming experience in Python or another programming language such as C, C++ or Matlab. Since Python will be the programming language used throughout this workshop, this lecture will teach the basics needed to be able to actively participate in all following sessions.

Wouter Klijn completed a MSc in Artificial Intelligence from the University of Groningen in the Netherlands. His Master thesis was on the information content of cell species in a 3-layer model of a cortical micro-column. After his studies, he worked in the software industry and research, focusing on big data real-time streaming systems and development of complex processing pipelines. He currently works at the Simulation Laboratory Neuroscience, Forschungzentrum Jülich as a PhD candidate with Professor Abigail Morrison. His research interests include the dynamics of large neural networks with a 2D spatial structure, and the design and operational aspects of big data science pipelines.

Lecture title: Scientific computing in Python part I  & II

This lecture will introduce the students with the basic scientific computing tools available in a Python environment. This includes the creation of and interaction with specialized data structures in the NumPy package, and the usage of these structures in the scientific operations as provided by SciPy. Additionally, visualization with the Matplotlib package will be shown. If time permits, high performance data storage with HDF5, or communication via MPI4Py will be discussed.

Susanne Kunkel works as a postdoctoral researcher at the Norwegian University of Life Sciences. She received a Diploma in Bioinformatics from the University of Jena, Germany and then started her doctoral research in Neuroinformatics at the Bernstein Center Freiburg. She received her doctoral degree (Dr. rer. nat.) from the University of Freiburg in 2015. Before moving to Norway, she worked at the Jülich Research Centre, Germany and at KTH Stockholm, Sweden. Susanne is one of the core developers of the NEST Simulator.

Lecture title: Getting started with NEST

In this lecture, I will introduce the NEST simulator (www.nest-simulator.org), which allows computational neuroscientists to investigate spiking neuronal network models in computer simulations. NEST focuses on the dynamics, size and structure of neuronal systems rather than on the exact morphology of individual neurons. The simulator is suitable for a wide range of applications: It enables distributed simulations of large-scale networks on high-performance computers but it also allows for interactive sessions on laptops using OpenMP threads.

Alberto Madonna is a software engineer within the HPC Operations unit at the Swiss National Supercomputing Centre (CSCS), currently working on container technologies for HPC environments and high-level/high-performance programming frameworks. His areas of expertise include the development of scientific and engineering analysis software, and GPU programming. Before joining CSCS, he was involved in developing data analysis software for parallel and low power architectures.

Lecture title: Introduction to High-Performance Computing 

This lecture will introduce the participants to the basics of High-Performance Computing (HPC). After an introduction to HPC system architectures, the participants will learn how such systems are operated and used by scientists for their research. Supercomputers are usually used in batch mode, which means that they are not used as interactively like standard computers or notebooks, but that a scheduler decides which job starts at which point in time and on which compute nodes to maximize the overall system usage.

Jan Meinke is a staff scientist at the Jülich Supercomputing Centre (JSC) and a member of the Simulation Laboratory Biology. He received his PhD in Physics in 2002 from Michigan State University and has been working at Forschungszentrum Jülich since 2005. His research interests include protein folding and finding ways to make efficient use of HPC hardware for solving scientific problems. He is a codeveloper of ProFASi, a Monte Carlo package for protein folding and aggregation simulations, and has been teaching Scientific Python courses since 2011.

Lecture title: Introduction to parallel computing part I & II

This lecture will introduce the participants to parallel computing using Python. After an introduction to parallel computing in general, including guidelines for choosing the right parts for parallelization based on profiling, the Python module numba will be introduced for sharedmemory parallelism and efficient singlenode implementations. In the second part of the lecture, the basic concepts and most important functions of MPI are introduced and then applied to some examples. MPI (Message Passing Interface) is the most commonly used distributed memory parallel programming model. Python supports MPI with mpi4py. The lecture will alternate between short presentations and hands-on exercises using Jupyter notebooks to enhance the learning experience and maximize the retention of the material.

Abigail Morrison is the group leader of “Computation in Neural Circuits” at INM- 6/IAS-6, Forschungszentrum Jülich (Germany), the leader of the “Simulation Laboratory Neuroscience” at the Jülich Supercomputing Centre and a professor at the Ruhr University of Bochum, Germany. She holds a master’s degree in artificial intelligence and received her PhD in computational neuroscience in 2006 from the University of Freiburg, Germany. Between 2006 and 2009 she was a scientific researcher at the RIKEN Brain Science Institute in Wako-Shi, Japan; she subsequently held a junior professorship at the University of Freiburg, Germany, as well as a group leadership at the Bernstein Center Freiburg, Germany from 2009 to 2012. Her research interests include learning, adaptation and self-organization in spiking neural networks, dynamics and computation in neurodegenerative diseases, and high-performance simulation technology.

Lecture title: Welcome & introduction by the Scientific Chair

In this introductory session, an overview of the workshop programme will be given by the Scientific Chair, Prof. Abigail Morrison. All tools and applications presented are part of the High Performance Analytics and Computing Platform of the Human Brain Project, which will also be briefly presented.

Lecture title: Introduction to the focus exercises 

With this session, we would like to give the students the opportunity to deepen their knowledge in the workshop topics they are most interested in. The students can choose from a set of experiments and exercises with different topics and varying difficulty that they can work on alone or in small groups, e.g. creating strong and weak scaling plots for a network, or to investigate how the scaling plot look like if conductancebased instead of current-based neurons are used. They can also bring own small projects or have a look at the hands-on parts of the earlier parts of the workshops to play around with them again. The workshop speakers and additional tutors from the SimLab Neuroscience are available to support the students with the exercises and mini-projects chosen by them.

Andreas Müller holds a bachelor in Scientific Programming and a master in Technomathematics. Already during his postgraduate studies, he worked at the Jülich Supercomputing Centre (JSC), Forschungszentrum Jülich and developed image processing software for neuroscientists on campus, that runs on the supercomputers at JSC. After obtaining his master’s degree in 2017, he moved to the division High Performance Computing in Neuroscience and began working on software for visualization and statistical analysis of neuroscientific graph models.

Lecture title: Guided tour to the supercomputing facilities at Jülich Supercomputing Centre

In this session, the participants will get an overview of JSC's supercomputers. They will learn how supercomputers can be compared with each other and what one needs to consider when writing code to run on them.

Lena Oden recently became a Junior Professor for Computer Architecture at the FernUniversität Hagen. Before that, she worked as a postdoctoral researcher at the Forschungszentrum Jülich and at Argonne National Laboratory in the USA. She received her PhD in Computer Science from the Ruprecht-Karls-Universität Heidelberg and a Diploma in Electrical Engineering from RWTH Aachen. During her PhD, she worked at the Fraunhofer Institute for Industrial Mathematics. Her main research areas are Computer Architectures and Runtime Systems for HPC.

Lecture title: HPC data management 

This lecture will give a short introduction into parallel file systems on HPC. We will discuss do’s and don'ts for handling large data sets, like avoiding copying data around or creating too many small files. We will discuss advantages and problems with parallel IO. We will also give a short introduction into HDF5, which allows an easy handling of large scientific data sets. We will use real examples from the Human Brain Project, in which optimizing IO helped to improve the performance of our applications.

Alexander Peyser holds bachelor degrees in anthropology, computer engineering and biology, practiced as a professional software engineer for seven years and in 2011 received a PhD in biophysics and neuroscience. He is a senior scientist and deputy scientific lead for the SimLab Neuroscience in the Jülich Supercomputing Centre. He is currently involved in the development of several simulation software packages for HPC (Arbor, NEST, HPC-for-TVB), HPC infrastructure for neuroscience and machine learning applications as well as large-scale, mathematically robust multiscale neuroscientific simulation.

Lecture title: Closing session

An overview of the material covered and a look forward to new technologies, applications and computer resources on the horizon. Where do we go from here?

Benjamin Weyers

is currently PostDoc at the Virtual Reality and Immersive Visualization group at RWTH Aachen University. He received his PhD in 2011 at the University of Duisburg-Essen and joined RWTH in 2013. He is interested in the development and research on interactive analysis methods for abstract and scientific data using immersive systems as well as the integration of VR and AR into the control of technical systems for the support of human user in semi-automated control scenarios.

Lecture title: Interactive visual data analysis

Interactive visual analysis tools become more and more an integral part of various data analysis processes in neuroscience. These tools are used for the exploration of small, medium or large data sets, to make neuroscientific phenomena visible for teaching and education, or for communication purposes in regard of presenting research results to the scientific community in form of publications or to the public. This session will present a broad overview of what such tools are capable of, how they can be used and how the right tool can be found for a specific research question. Therefore, we will give a brief introduction into visualization methods followed by the presentation of various tools and use cases from the domains of computational and experimental neurosciences.