Conference Presentation (Invited)

FZJ-2023-00582

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png On concepts, correctness, and performance of network simulations

Senk, J. (Corresponding author)FZJ*

2022

Abstract: The development of large-scale neuronal network models is an iterative process and best described by a loop linking the conceptual description with its algorithmic implementation. Models are informed by experimental data and insights from theory; their simulated dynamics needs to be validated against experimentally observed activity. The inherent interdisciplinarity and intricacy of this endeavor make it prone to pitfalls that complicate the comprehensibility and reproducibility of each step. Addressing shortcomings in model descriptions, we review how connectivity is specified in published models and propose unified connectivity concepts and guidelines including a graphical notation for network diagrams. Furthermore, we focus on the challenge to compare simulations of a cortical microcircuit model with respect to correctness and performance across different simulation technologies; literature shows a race for the fastest and most energy efficient simulation. We also showcase a conceptual workflow for performance benchmarking of the simulator NEST together with the benchmarking framework beNNch as a reference implementation. Finally, we introduce NNMT, a toolbox for mean-field based analysis methods of neuronal network models. The presented approaches aim at raising awareness to common difficulties and fostering a sustainable and collaborative modeling culture.

Note: Session: Tools and formats for large scale network modelling

---

Contributing Institute(s):

1. Computational and Systems Neuroscience (INM-6)
2. Theoretical Neuroscience (IAS-6)
3. Jara-Institut Brain structure-function relationships (INM-10)

Research Program(s):

1. 5231 - Neuroscientific Foundations (POF4-523) (POF4-523)
2. 5234 - Emerging NC Architectures (POF4-523) (POF4-523)
3. 5235 - Digitization of Neuroscience and User-Community Building (POF4-523) (POF4-523)
4. HBP SGA1 - Human Brain Project Specific Grant Agreement 1 (720270) (720270)
5. HBP SGA2 - Human Brain Project Specific Grant Agreement 2 (785907) (785907)
6. HBP SGA3 - Human Brain Project Specific Grant Agreement 3 (945539) (945539)
7. DEEP-EST - DEEP - Extreme Scale Technologies (754304) (754304)
8. GRK 2041 - GRK 2041: Modell Romantik. Variation - Reichweite - Aktualität (250805958) (250805958)
9. DFG project 347572269 - Heterogenität von Zytoarchitektur, Chemoarchitektur und Konnektivität in einem großskaligen Computermodell der menschlichen Großhirnrinde (347572269) (347572269)
10. ACA - Advanced Computing Architectures (SO-092) (SO-092)
11. neuroIC001 - NeuroModelingTalk (NMT) - Approaching the complexity barrier in neuroscientific modeling (EXS-SF-neuroIC001) (EXS-SF-neuroIC001)
12. MetaMoSim - Generic metadata management for reproducible high-performance-computing simulation workflows - MetaMoSim (ZT-I-PF-3-026) (ZT-I-PF-3-026)
13. JL SMHB - Joint Lab Supercomputing and Modeling for the Human Brain (JL SMHB-2021-2027) (JL SMHB-2021-2027)
14. MSNN - Theory of multi-scale neuronal networks (HGF-SMHB-2014-2018) (HGF-SMHB-2014-2018)

Appears in the scientific report 2023

---