High Performance Computing (HPC) Resources for Parallel Simulations and Data Analysis: NSG and HPAC

Saturday, 11 November, 2017, Time: 8:30am - 12:30pm 
Satellite Workshop, SfN 2017, Washington D.C., USA

Location: Downtown Washington D.C. (will be informed to attendees)

Registration required, deadline Friday, October 27, 2017 – register here.

WORKSHOP ORGANIZERS

Amit Majumdar, Subhashini Sivagnanam, Kenneth Yoshimoto
San Diego Supercomputer Center, University of California San Diego, La Jolla, CA, USA

Ted Carnevale
Department of Neuroscience, Yale University, New Haven, CT, USA

Alexander Peyser
Institute for Advanced Simulation (IAS), Jülich Supercomputing Centre, Germany

Abstract of the workshop:

Need free, easy access to HPC resources to run widely used neural simulators or connectome analysis tools? Interested in new, free HPC tools developed by the HBP, including hardware, simulation and analytics software?

The Neuroscience Gateway (NSG) project and the High Performance Analytics and Computing (HPAC) of the Human Brain Project will host a joint satellite workshop at the Society for Neuroscience (SfN) 2017 annual meeting in Washington D.C. Workshop presenters are neuroscientists who are involved in computational neuroscience research and education as well as developers of tools (such as NEST, NEURON) used in computational neuroscience.

Title of Talks and Names of Speakers:

9:00 – 9:20 The Neuroscience Gateway – enabling large scale simulation and data processing for computational neuroscience, Amit Majumdar, Subhashini Sivagnanam, Kenneth Yoshimoto San Diego Supercomputer Center, UCSD; Ted Carnevale, Department of Neuroscience, Yale University, USA

9:20 – 9:30 Computational ecosystem for neuroscience, Alexander Peyser, Institute for Advanced Simulation (IAS), Jülich Supercomputing Centre, Germany

9:30 – 10:00 Modeling approach to whole brain dynamics and simulations with The Virtual Brain (TVB), Andreas Spiegler, Institut de Neurosciences des Systèmes (INS), Aix-Marseille Université (AMU), France

10:00 – 10:30 Training modules for data intensive neuroscience learning and research, Satish Nair, Electrical and Computer Engineering, University of Missouri, USA

10:30 – 10:40 Break

10:40 – 11:10 NEST, brain building blocks integrated with theory, Markus Diesmann, Institute of Neuroscience and Medicine Computational and Systems Neuroscience, Institute for Advanced Simulations & JARA Brain Institute, Research Center Jülich; Department of Psychiatry, Psychotherapy and Psychosomatics; Department of Physics, RWTH Aachen University, Germany

11:10 – 11:40 Reconstruction of a full-scale model of the rat hippocampus CA1, Armando Romani, Blue Brain Project, EPFL, Switzerland

11:40 – 12:10 High performance computing for analysis and design of peripheral nerve stimulation, Nicole A. Pelot, Warren M. Grill, Department of Biomedical Engineering, Duke University, USA

12:10 – 12:30 Lunch and end of workshop

 

SCHEDULE OF EVENTS


9:00 - 9:20 

The Neuroscience Gateway – enabling large scale simulation and data processing for computational neuroscience

Amit Majumdar, Subhashini Sivagnanam, Kenneth Yoshimoto
San Diego Supercomputer Center, University of California San Diego, La Jolla, CA, USA

Ted Carnevale
Department of Neuroscience, Yale University, New Haven, CT, USA

We will provide a brief overview of the Neuroscience Gateway, including all the tools, software and pipelines that are provided via NSG on supercomputing resources. We will discuss how NSG has evolved over the past years where experimental and cognitive neuroscientists are using NSG for their research. NSG has also become a platform where developers of neuroscience software and tools are releasing their product for the wider neuroscience community. NSG is also being used for educational projects. From the NSG perspective we will discuss the opportunities, ideas and interest to interact with other projects in a collaborative way to serve the broader neuroscience community world-wide.


9:20 - 9:30 

Computational ecosystem for neuroscience simulation 

Alexander Peyser
Institute for Advanced Simulation (IAS), Jülich Supercomputing Centre, Germany

This talk will provide an overview of computational resources that are being used for large scale neuronal simulation. It will discuss the current and future HPC architectures of interest to computational neuroscientists.


9:30 - 10:00 

Modeling approach to whole brain dynamics and simulations with The Virtual Brain (TVB)

Andreas Spiegler
Institut de Neurosciences des Systèmes (INS), Aix-Marseille Université (AMU), France

How do large-scale brain dynamics arise from the anatomical structure of brains and support brain function and dysfunctions in movement, cognition, and perception? The necessary systematic explorations can only be performed in silico with the help of HPC. This talk addresses the questions of how to simulate (and analyze) whole brain dynamics. The neuroinformatics platform The Virtual Brain (TVB) is tailored to large-scale brain modeling. TVB will be presented and the technical requirements and challenges will be discussed. Examples of applications will demonstrate that HPC makes whole-brain simulations and parameter explorations feasible. 


10:00 - 10:30 

Training modules for data intensive neuroscience learning and research 

Satish Nair
Electrical and Computer Engineering, University of Missouri, USA

We will discuss a NSF funded project that will develop cyberinfrastructure-based training modules that advance the existing training methods used for learning and research in data-intensive neuroscience communities. The project outcomes will enhance research into our understanding of both normal and abnormal brains, contributing to NSF's mission of advancing progress in both science and health. The project activities will address important gaps in existing training methods that arise because neuroscience research and education activities are increasingly becoming data-intensive. NSG will be used as a dissemination platform for some of the modules and tools developed as a part of this project. 


10:30 - 10:40

Break

10:40 - 11:10 

NEST, brain building blocks integrated with theory  

Markus Diesmann
Institute of Neuroscience and Medicine Computational and Systems Neuroscience, Institute for Advanced Simulations & JARA Brain Institute, Research Center Jülich; Department of Psychiatry, Psychotherapy and Psychosomatics; Department of Physics, RWTH Aachen University, Germany

The simulation engine NEST is widely used in the community for the investigation of large-scale spiking neuronal network models. The tool enables the study of networks at their natural density of connectivity.

The talk demonstrates the use of NEST for the creation of network models as reusable building blocks. In this way NEST supports the reproducibility in computational neuroscience and helps the field to overcome a complexity barrier: It seems plausible that future progress in our field depends on the capability of neuroscientists to construct more complete models of brain function from already existing and validated modules.

The next part of the talk reports on the recent inclusion of gap junction interaction [1] as a further biophysical mechanism, on improvements of the performance on modern computer architectures ranging from laptops to the largest supercomputers [2,3], and on a language (NESTML) for the specification of point-neuron models in the terms of the problem domain [4].

While computational neuroscience has made progress by establishing common model description languages and simulation platforms for the abstraction level of networks of spiking neurons, the consolidation is less advanced for more abstract models, such as binary or rate-based models for individual neurons or populations thereof. The last part of the talk introduces the technology [5] for simulating networks of continuously interacting units in NEST. This feature enables theoretical neuroscientists working above the level of abstraction of individual spiking neurons to simulate and publish their models on the basis of a widely used platform. Furthermore, the level of abstraction in a complex network model can be changed by modification of a few lines in the formal model description. Finally, the technology is now in place to construct models with components at different levels of abstraction.

The open development of NEST is guided by the NEST Initiative (www.nest-initiative.org). Partial funding comes from the European Human Brain Project through EU grants 604102 and 720270.

[1] Hahne J, Helias M, Kunkel S, Igarashi J, Bolten M, Frommer A, Diesmann M (2015) Frontiers in Neuroinformatics 9:22

[2] Kunkel S, Schmidt M, Eppler JM, Plesser HE, Masumoto G, Igarashi J, Ishii S, Fukai T, Morrison A, Diesmann M, Helias M (2014) Frontiers in Neuroinformatics 8:78

[3] Ippen T, Eppler JM, Plesser HE, Diesmann M (2017) Frontiers in Neuroinformatics 11:30

[4] Plotnikov D, Blundell I, Ippen T, Eppler JM, Morrison A, Rumpe B (2016) “Modellierung 2016” Conference 93:108

[5] Hahne J, Dahmen D, Schuecker J, Frommer A, Bolten M, Helias M, Diesmann M (2017) Frontiers in Neuroinformatics 11:34.


11:10 - 11:40 

Reconstruction of a full-scale model of the rat hippocampus CA1

Armando Romani
Blue Brain Project, EPFL, Switzerland

We present a first draft model of the rat hippocampus CA1 algorithmically reconstructed according to the technology developed by the Blue Brain Project (BBP) and made publicly available through the Human Brain Project (HBP) Brain Simulation Platform (BSP).

In brief, the reconstruction process began by placing a series of morphological reconstructions in the hippocampal volume defined by publically available atlases and according to experimentally measured densities and composition. A connectome was predicted as previously described (Reimann et al 2015), using biological data on bouton densities and synapses per connection. Electrical models of neurons and synaptic physiology were constrained by electrophysiological recordings and published data. Finally, additional datasets were used to validate each of the reconstruction steps, and emerging properties of the complete model.

The hippocampus model aims to be part of a larger community initiative involving groups beyond the HBP, which will periodically refine the model and make regular public releases. The reconstruction represents a resource for the community to integrate experimental data, perform in silico experiments, and test hypotheses on hippocampal function. 


11:40 - 12:10 

High performance computing for analysis and design of peripheral nerve stimulation

Nicole A. Pelot, Warren M. Grill
Department of Biomedical Engineering, Duke University, USA 

There is rapidly growing interest in peripheral nerve stimulation to treat a broad range of diseases. To design effective therapies, we must understand the biophysical mechanisms of action, as well as the nerve response patterns across a large, multi-dimensional parameter space. Such understanding can be gleaned through brute force parameter sweeps or through optimization algorithms working towards a desired target outcome. Both types of studies are feasible through massive parallelization of the simulations on a computer cluster. In this presentation, I will present findings using both approaches. First, I will summarize activation and block response patterns to kilohertz frequency signals across stimulation amplitudes, frequencies, fiber diameters, and electrode designs. This approach provides an overview of the range of possible neural responses across typical clinical parameters. Second, I will present optimization of monopolar and bipolar pulses to elicit action potentials with minimum energy using particle swarm optimization. The results demonstrate the power of computational modeling for analysis and design of peripheral nerve stimulation, and highlight the importance of these approaches to advance bioelectronics medicine.

Acknowledgments: This work was supported by the National Institutes of Health [OT2 OD025340]; the Natural Sciences and Engineering Research Council of Canada [PGS D3-437918-2013]; and the Myra & William Waldo Boone Fellowship from the Graduate School of Duke University. 


12:10 - 12:30

Lunch and end of workshop

Starts 11 Nov 2017 08:30
Ends 11 Nov 2017 12:00
US/Eastern